Rapid Discovery of CD38 Inhibitor via DNA-Encoded Natural Product Library Screening

Abstract

CD38 is a multifunctional enzyme that plays a pivotal role in NAD⁺ metabolism and calcium signaling, and its abnormal activity is closely associated with multiple myeloma, age-related metabolic decline, neurodegenerative diseases, and other disorders. Although monoclonal antibodies such as daratumumab have been approved for clinical application, their inherent limitations necessitate the development of novel small-molecule CD38 inhibitors. In this study, we employed DNA-encoded library (DEL) technology for the high-throughput screening of CD38 inhibitors, using a DEL library containing more than 100,000 unique compounds to screen against recombinant human CD38. A total of 1043 enriched compounds were initially identified, and after rigorous validation and screening to exclude non-specific binding and previously reported active compounds, eight hit compounds with diverse chemical scaffolds were obtained, among which Fenbendazole—a clinically approved antiparasitic drug—was included. Surface plasmon resonance (SPR) assays confirmed the direct binding of these hit compounds to CD38, with dissociation constants (KD) ranging from 7.74 × 10⁻⁵ M to 2.15 × 10⁻⁴ M. Fluorescence-based enzymatic activity assays demonstrated that these compounds exert dose-dependent inhibitory effects on both the hydrolase (with ε-NAD as substrate) and cyclase (with NGD as substrate) activities of CD38. Further structure–activity relationship (SAR) analysis of Fenbendazole analogues revealed the critical structural features that regulate CD38 inhibitory potency, and Flubendazole was found to exhibit excellent inhibitory activity, with an IC₅₀ of 14.78 ± 4.21 μM against CD38 hydrolase and 26.31 ± 3.40 μM against cyclase. Molecular docking and 100 ns molecular dynamics (MD) simulations further elucidated the molecular mechanism of CD38 inhibition by lead compounds, confirming that van der Waals interactions are the main driving force for the binding of small-molecule ligands to CD38, with conserved aromatic residues in the active site mediating ligand recognition. This study validates DEL technology as an efficient and reliable platform for the discovery of CD38 inhibitors, and the identified lead compounds—especially Fenbendazole and its analog Flubendazole—provide valuable molecular scaffolds for the further structural optimization of CD38 inhibitors. These findings lay a solid foundation for the development of novel therapeutic agents for the treatment of CD38-associated diseases.

1. Introduction

CD38 is a type II transmembrane glycoprotein with multiple enzymatic activities [1,2] including ADP-ribosyl cyclase, cyclic ADP-ribose hydrolase, and NAD⁺ glycohydrolase activities [3]. It plays a crucial role in regulating intracellular and extracellular levels of NAD⁺, a key coenzyme involved in numerous metabolic reactions, as well as cADPR, an important second messenger in calcium signaling [4,5,6,7,8,9,10,11]. Aberrant expression or overactivation of CD38 has been implicated in the pathogenesis of various diseases [12,13,14]. For example, high CD38 expression is a hallmark of multiple myeloma (MM) cells, where it promotes tumor growth by depleting NAD⁺ and suppressing anti-tumor immunity [15]. In addition, CD38-mediated NAD⁺ depletion is associated with age-related metabolic decline and neurodegenerative diseases, making it a potential target for anti-aging interventions.

To date, several CD38 inhibitors have been developed, including small molecules and monoclonal antibodies, with representative examples depicted in Figure 1 [16,17,18,19]. Evolving from the initial flavonoids and NAD substrate analogues, the field saw a significant advancement in 2015 when GSK reported two series of CD38 inhibitors. Compounds such as 1bf and 78c exhibited improved potency and suitable pharmacokinetic profiles for in vivo animal studies [20,21]. The X-ray crystal structure of the 1bf-bound complex and 78c-ADPR adduct in complex with CD38 was resolved, providing valuable insights into the catalytic determinants for substrate recognition by CD38 and the binding modes of these two representative inhibitors. Subsequent optimization efforts led to the identification of promising drug candidates, such as RBN013209, developed by Ribon Therapeutics. This compound demonstrated antitumor activity in a melanoma model when used in combination with a checkpoint inhibitor [20]. More recently, Mitobridge introduced MK-0159, a CD38 enzymatic inhibitor that showed marked cardioprotective effects during cardiac ischemia/reperfusion (I/R) injury [22]. Through a scaffold hopping strategy, the potency and overall drug-like properties of MK-0159 were significantly enhanced.

Figure 1.

Representative CD38 inhibitors.

CD38 inhibitors identified to date exhibit distinct chemical structural features, including core scaffolds such as quinoline carboxamides (e.g., 4-amino-8-quinoline carboxamides), indole-7-carboxamides, pyrrolo[2,3-b]pyridines (e.g., MK-0159), and thiazoloquin(az)olin(on)es [23,24,25]. Key structural elements include substituted aromatic rings (e.g., dichloro-, fluoro-, trifluoromethyl-substituted benzyl groups) at the 4-position of quinoline or 5-position of indole, essential 8-carboxamide moieties for quinoline derivatives (critical for hydrogen bonding with CD38 residues like Glu¹⁴⁶ and Asp¹⁵⁵), and specific heterocyclic substituents (e.g., imidazole, thiazole) that maintain inhibitory potency [26,27]. Additionally, favorable analogues often feature trans-geometry cyclohexylamine side chains and electron-withdrawing groups (e.g., nitro, cyano) at specific positions to enhance binding affinity while showing high selectivity for CD38 over homologs like CD73 and CD157 [28,29,30,31,32]. Monoclonal antibodies such as daratumumab have been approved for the treatment of MM [33,34,35], but they suffer from limitations such as high production costs and potential immunogenicity [36,37,38,39,40,41]. Small-molecule inhibitors, on the other hand, offer the advantages of oral bioavailability, lower cost, and broader applicability [42,43,44]. Flavonoids, a class of natural products, have been reported to exhibit CD38 inhibitory activity, highlighting the potential of this structural class in CD38 inhibitor development [45,46]. However, the discovery of novel small-molecule CD38 inhibitors with improved potency, selectivity, and druggability remains a major challenge [19,47].

DNA-encoded chemical library (DEL) technology has emerged as a game-changing platform for high-throughput screening in drug discovery [48,49,50]. Unlike traditional high-throughput screening (HTS), which requires the individual handling of compounds, DEL technology encodes each compound with a unique DNA barcode, allowing millions to billions of compounds to be screened in a single tube [51,52]. The core advantages of DEL include significantly reduced screening costs with minimal target protein consumption and the rapid identification of hit compounds through high-throughput sequencing and bioinformatics analysis [53,54]. DEL has been successfully applied to the discovery of inhibitors for various therapeutic targets, including enzymes, receptors, and protein–protein interaction interfaces [55,56,57].

Recent advances in structure-based drug design (SBDD), particularly through the integration of molecular docking with multidisciplinary approaches, have significantly accelerated anticancer drug discovery [58,59]. For instance, a novel thiazole-chalcone/sulfonamide hybrid (compound 7) was rationally designed and synthesized as a dual inhibitor of tubulin polymerization and carbonic anhydrase IX, with molecular docking confirming its strong interactions with both targets, while computational ADMET and DFT analyses validated its drug-like properties and stability [60]. In another case, the natural product α-hederin was identified as a potent and selective JAK1/JAK2 inhibitor through virtual screening and docking, and its mechanism was rigorously confirmed by transcriptomics, biochemical assays, and in vivo models, demonstrating efficacy against ovarian cancer with minimal toxicity [61]. These studies underscore how the synergy of docking with experimental validation, computational profiling, and mechanistic exploration accelerates the discovery of effective, multi-target anticancer agents.

Here, we present an integrated strategy that combines the power of DNA-encoded library (DEL) technology with structure-based drug design (SBDD) workflows. We employed DEL screening to rapidly identify hit compounds against human CD38, addressing the limitations of traditional screening. Subsequently, we applied SBDD principles—including molecular docking, structural analysis, and dynamic simulations—to characterize target–ligand interactions and guide the optimization of lead candidates. This dual approach ensures both efficient hit discovery and rational design, ultimately paving the way for the development of novel CD38 inhibitors.

2. Results

2.1. DEL Screening and Identification of Hit Compounds

To discover novel small-molecule binders of human CD38, we performed an affinity selection screen using a DNA-encoded library (DEL) comprising over 0.1 million compounds. For structural context, the binding site of CD38 is depicted in complex with a known small-molecule inhibitor 78c [23] (Figure 2a). The recombinant CD38 protein was immobilized and incubated with the DEL pool. Following stringent affinity selection, washing, and elution with 78c steps, the DNA barcodes of the bound compounds were recovered, amplified, and decoded via high-throughput sequencing. The screening pipeline, from library construction to hit identification, is summarized in Figure 2b.

Figure 2.

Structural Insights and Screening Pipeline for CD38-Targeted Compound Discovery. (a) Structure of CD38 in complex with the small-molecule inhibitor 78c and the ADPR adduct (PDB: 8DOM). CD38 is depicted as a cartoon in slate, with the N-lobe and C-lobe labeled. The small-molecule inhibitor is shown in stick representation. (b) General workflow of DNA-Encoded Library (DEL) screening. A DEL is constructed by conjugating oligonucleotide tags to chemical building blocks through iterative split-and-pool synthesis. The immobilized target protein is then incubated with the DEL for screening. After washing away non-specifically bound compounds, the enriched target-binding molecules are eluted, and their DNA barcodes are extracted. These barcodes are amplified by PCR, sequenced via high-throughput next-generation sequencing (NGS), and then bioinformatically matched against a barcode-compound database to identify hit compounds. Promising candidates are subsequently synthesized and validated as lead compounds. (c) Table of decoded small molecules ranked by enrichment fold (from highest to lowest). The small molecules selected for further validation were assigned numerical codes and annotated in the table for clarity. (d) Chemical structure diagrams of selected small molecules chosen for further validation, labeled with their assigned numerical codes.

Comparative analysis against a control resin identified 1043 compounds that exhibited significant enrichment (enrichment ratio > 1). From this enriched pool, we prioritized 21 potential hit compounds for downstream validation based on two stringent criteria: high sequencing read counts (>2.5) indicative of robust enrichment, and diverse core chemical scaffolds. The decoded small molecules are ranked by their enrichment fold in Figure 2c, with the 21 selected hits clearly annotated. To refine the hit list, an empirical filter was applied to exclude promiscuous small molecules known to exhibit strong, non-specific adsorption. Such compounds are frequently observed across independent DEL screens against various protein targets and are considered undesirable false positives. In Figure 2c, these non-specifically binding compounds are indicated in gray and were omitted from further validation. Furthermore, known flavonoid-based CD38 inhibitors, such as quercetin and luteolin, were deliberately excluded from the final selection, as their activity has been previously documented in the literature [46]. Ultimately, eight compounds were selected from the candidate pool for further validation and are hereafter referred to by new numerical codes (Figure 2c). This final curated set encompasses diverse pharmacologically active scaffolds, including a fluoroquinolone (Sparfloxacin, 2), a peripherally-acting μ-opioid receptor antagonist (Alvimopan, 8), a flavonoid glycoside (Hyperoside, 3), an anthraquinone derivative (Diacerein, 5), a xanthine oxidase inhibitor (Topiroxostat, 6), a purine analogue immunosuppressant (Azathioprine, 7), a benzimidazole anthelmintic (Fenbendazole, 4), and a diterpenoid (Oridonin, 1), demonstrating the ability of the DEL to interrogate broad and privileged chemical spaces (Figure 2d).

2.2. Binding Affinity of Hit Compounds to CD38

Following the identification of eight potential hit compounds through DEL screening, the next critical step was to verify their direct binding to CD38 and quantify their binding affinities. Surface plasmon resonance (SPR) was employed for this purpose, as it enables the real-time, label-free analysis of biomolecular interactions. Since Oridonin (1) did not exhibit binding, the cutoff for the enrichment fold in our NGS data analysis was adjusted accordingly to ensure stringency. Other compounds exhibited concentration-dependent binding to the extracellular domain of human CD38, as evidenced by the dose–responsive sensorgrams (Figure 3, upper panels). Equilibrium binding isotherms fitted to a 1:1 binding model yielded dissociation constants (K_D) ranging from 6.66 × 10⁻⁴ M to 4.67 × 10⁻⁶ M (Figure 3, lower panels). The majority of the compounds showed moderate binding affinity, with K_D values clustering within the 10⁻⁴ M range (e.g., approximately 2.52 × 10⁻⁴ M for Fenbendazole [compound 4]), while Topiroxostat and Alvimopan (compound 6 and 8, respectively) exhibited a slightly higher affinity of approximately 7 × 10⁻⁵ M (Figure 3d,f,h), while 78c was determined at 7.06 × 10⁻⁵ M (Figure S1).

Figure 3.

SPR binding analysis of selected compounds to human CD38 extracellular protein. The upper panels (a–h) depict the binding sensorgrams of comp 1–8# to soluble CD38 at varying concentrations, with the respective concentration of each analyte labeled alongside and represented by a corresponding curve color. The lower panels show the corresponding equilibrium binding isotherms fitted to determine affinity, with the calculated K_D value annotated on each plot. The chemical structure of each compound is also provided adjacent to its respective data to facilitate structure–activity interpretation.

It is worth noting that Fenbendazole (4) exhibited slow binding kinetics, suggesting that it may form a stable complex with the target, despite its relatively low apparent affinity in the steady-state simulation. These SPR results confirm that the hit compounds identified from DEL screening are indeed capable of directly interacting with CD38, laying the foundation for the subsequent evaluation of their inhibitory effects on CD38 enzymatic activities.

2.3. Inhibitory Activity of Hit Compounds on CD38 Enzymatic Functions

Having confirmed the direct binding of the eight hit compounds to CD38, we next sought to determine whether these bindings translated into functional inhibition of CD38’s enzymatic activities. CD38 exhibits two major enzymatic activities: NAD⁺ hydrolase (cleaving NAD⁺ to ADPR and nicotinamide) and cyclase (converting NAD⁺ to cyclic ADP-ribose, cADPR) [2,20]. We evaluated the inhibitory effects of hit compounds on both activities using fluorescence-based assays with ε-NAD (hydrolase substrate) and NGD (a cyclase substrate alternative for facile plate-reader detection), respectively [21,22].

In the assessment of hydrolase activity, compounds 3 (Hyperoside), 4 (Fenbendazole), 6 (Topiroxostat), and 7 (Azathioprine) exhibited a certain degree of inhibition on product generation (Figure 4a). This inhibition was observable at both 10 min (Figure 4b) and 30 min (Figure 4c), compared to the NC control (Figure 4a–c). Fenbendazole and Topiroxostat showed the strongest inhibitory effects, reducing hydrolytic product formation by >50% at both 10 min and 30 min.

Figure 4.

Inhibitory Activity of Hit Compounds on CD38 Enzymatic Functions. (a) Enzyme activity inhibition curves of the tested compounds at 100 μM against CD38 hydrolase. (b,c) Bar graphs quantifying the hydrolytic product levels after 10 min (b) and 30 min (c) of reaction across different experimental groups. (d) Enzyme activity inhibition curves of the tested compounds at 50 μM against CD38 cyclase. (e,f) Bar graphs quantifying the cyclized product levels after 10 min (e) and 30 min (f) of reaction across different groups. The data point for compound 6 was excluded from the analysis in (e) due to an anomalous signal.

Consistent with previous findings, cyclase activity was also inhibited. This was evidenced by the suppression of cGDPR production from NGD, with Fenbendazole and Topiroxostat showing the most potent dose-dependent effects (Figure 4d–f). Taken together, these results demonstrate that the hit compounds not only bind to CD38 but also act as potent dual inhibitors of CD38, suppressing both its hydrolase and cyclase activities through binding at the catalytic site. This dual inhibitory property is particularly valuable, as it addresses both the NAD⁺ depletion and calcium signaling dysregulation associated with CD38 overactivation in disease states.

2.4. Structural Diversity Screening of Fenbendazole Analogues for SAR Exploration

Among the eight hit compounds, Fenbendazole stood out due to its moderate binding affinity, potent dual inhibitory activity, and status as a clinically approved antiparasitic agent, which implies a known safety profile and potential for drug repurposing. To further explore the structure–activity relationship (SAR) of benzimidazole derivatives as CD38 inhibitors, we selected several Fenbendazole analogues with structural diversity at the R1 and R2 positions of the Fenbendazole scaffold (Figure 5a).

Figure 5.

Structural diversity of Fenbendazole-based benzimidazole analogues. (a) Chemical formula illustrates the structural variations at the R1 and R2 positions of the Fenbendazole scaffold among the selected benzimidazole derivatives. (b–j) Details of individual structures are provided.

The R1 position encompasses compounds with flexible side chains, including Albendazole (Figure 5b), Oxibendazole (Figure 5f), Parbendazole (Figure 5d), and Bendamustine (Figure 5j)—a bifunctional alkylating agent known for its antitumor and cytotoxic effects. In contrast, the R1 position also features compounds with rigid, predominantly aromatic side chains, such as Mebendazole (Figure 5c), Flubendazole (Figure 5g), Oxfendazole (Figure 5e), and Nocodazole (Figure 5h). Additionally, Thiabendazole (Figure 5i) represents a case where this position is unsubstituted. Structural diversity at the R2 position is relatively limited. Most of the compounds share a consistent methoxycarbamide group, as seen in Figure 4d. The only exceptions are Thiabendazole (Figure 5i), which features a thiazole ring at this position, and Bendamustine (Figure 5j), where the R2 group is a butyrate moiety. By systematically investigating these analogues, we aimed to identify critical structural features that contribute to CD38 binding affinity and inhibitory potency, providing guidance for the rational design of more effective CD38 inhibitors.

2.5. SPR Analysis of Binding Affinity for Fenbendazole Derivatives

To elucidate the impact of structural modifications at the R1 and R2 positions on CD38 binding, we performed SPR analysis on the selected Fenbendazole analogues. The results revealed distinct binding affinities among the analogues (Figure 6). Compounds bearing rigid aromatic R1 side chains, such as Oxfendazole (4-4), exhibited significantly higher affinity (K_D ~2.15 × 10⁻⁴ M) than those with flexible R1 side chains, such as Oxibendazole (4-5, K_D = 4.38 × 10⁻³ M) and Parbendazole (4-6, K_D = 6.42 × 10⁻⁴ M). This suggests that aromatic substituents at R1 enhance binding to CD38, likely via hydrophobic interactions with active-site residues (e.g., Trp¹⁸⁴, Phe²²⁹). Among the analogues, Bendamustine (4-9) showed the highest binding affinity (K_D = 1.31 × 10⁻⁴ M). These SPR results provide valuable insights into the SAR of Fenbendazole analogues, highlighting the importance of the R1 substituent in mediating CD38 binding.

Figure 6.

SPR binding analysis of Fenbendazole (4) derivatives to the extracellular domain of human CD38. Top panels (a–f) show the binding sensorgrams for each compound at the indicated concentrations, with each concentration represented by a distinct colored curve. Bottom panels display the corresponding equilibrium binding isotherms fitted to determine binding affinity; the calculated K_D value is annotated on each plot. The chemical structure of each analyzed compound is provided adjacent to its binding data to aid in structure–activity relationship analysis.

2.6. Inhibitory Activity of Fenbendazole Analogues

Building on the binding affinity data, we next evaluated the inhibitory activity of Fenbendazole analogues on CD38’s hydrolase and cyclase activities using the same fluorescence-based assays as before. Enzymatic assays confirmed that structural modifications at R1 and R2 modulate inhibitory potency (Figure 7). Oxibendazole (4-5) showed the strongest inhibition of the hydrolase activity of CD38 when tested at 100 μM (Figure 7a,b). However, during the cyclase activity assay, the background-subtracted signal of the Oxibendazole group was negative, likely due to the compound’s strong intrinsic fluorescence within the detection wavelength range (Figure 7c,d). This fluorescence interference compromised the reliability of the cyclase activity data. To ensure the clarity and accuracy of data presentation, Oxibendazole was excluded from the cyclase activity analysis. Consequently, no definitive conclusion can be drawn regarding Oxibendazole acting as a dual inhibitor of CD38 hydrolase and cyclase activities. Other analogues, such as Flubendazole (4-3) and Parbendazole (4-6), exhibited inhibitory effects comparable to those of Fenbendazole (4) on both CD38 hydrolase and cyclase activities (Figure 7). Notably, Flubendazole, with its difluorophenyl R1 group, exhibited an IC₅₀ of 14.78 ± 4.21 μM against hydrolase activity and 26.31 ± 3.40 μM against cyclase activity, making it a promising lead compound for further optimization (Figure 7e–h). The affinity of all compounds for CD38 was integrated (Table 1). The results, combined with the SPR data, provide a comprehensive SAR profile of Fenbendazole analogues, identifying key structural features that contribute to CD38 inhibitory potency.

Figure 7.

Inhibitory Activity of Fenbendazole Analogues on CD38 Enzymatic Functions. (a) Inhibition curves of the tested compounds (final concentration: 100 μM) against CD38 hydrolase activity using e-NAD as the substrate. (b) Quantification of hydrolytic product levels after 10 min of reaction across experimental groups. (c) Inhibition curves of the tested compounds (final concentration: 50 μM) against CD38 cyclase activity using NGD as the substrate. (d) Quantification of cyclized product levels after 10 min of reaction across groups. (e–g) Dose-dependent inhibition of CD38 hydrolase by Fenbendazole and Oxibendazole. Dose–response curve and corresponding IC₅₀ value for Fenbendazole (f) and Oxibendazole (h) against human CD38 hydrolase activity. Data points represent the mean ± SD of triplicate measurements. Calculated IC₅₀ values are indicated within each panel.

Table 1.

The affinity of all compounds for CD38.

Compound	KD(M)
78c	7.06 × 10−5
Sparfloxacin	6.66 × 10−4
Hyperoside	2.99 × 10−4
Fenbendazole	2.52 × 10−4
Diacerein	4.67 × 10−6
Topiroxostat	7.51 × 10−5
Azathioprine	4.99 × 10−4
Alvimopan	7.74 × 10−5
Oridonin	No binding
Oxfendazole	2.15 × 10−4
Oxibendazole	4.38 × 10−3
Parbendazole	6.42 × 10−4
Thiabendazole	2.24 × 10−3
Bendamustine	1.31 × 10−4
Albendazole	No binding

2.7. Structural and Energetic Basis of CD38 Inhibition by Small-Molecule Ligands

To elucidate the molecular mechanisms underlying CD38 inhibition, we integrated molecular docking with 100 ns molecular dynamics (MD) simulations to characterize the binding modes, conformational stability, and energetic determinants of three lead compounds—Topiroxostat, Fenbendazole, and Oxibendazole—at the CD38 active site, using an apo structure of CD38 as the receptor model [62], as seen in Figure 8. Initial docking revealed that all three ligands engaged core catalytic residues, with a conserved set of aromatic residues including TRP-125 and TRP-189 involved in ligand recognition, while forming ligand-specific interaction networks. The CD38–Topiroxostat complex formed hydrogen bonds with TRP-125 and THR-221, complemented by hydrophobic packing with ARG-127, TRP-125, and LEU-145 (Figure 8a–c). Fenbendazole was bound via hydrogen bonds with GLN-226 and TRP-125, accompanied by π-mediated hydrophobic interactions (Figure 8d–f). Oxibendazole was stabilized by hydrogen bonds with SER-193 and ASP-156, together with hydrophobic contacts involving TRP-189, TRP-125, and LEU-145 (Figure 8g–i).

Figure 8.

Schematic diagram of in silico predicted key CD38–ligand interactions at the binding pocket. Cartoon representations of CD38 in complex with Topiroxostat (a), Fenbendazole (d), and Oxibendazole (g). The protein is depicted as a rainbow-colored cartoon, while the small molecules are shown as stick-and-ball models. In silico predicted three-dimensional representations showcasing the binding modes of CD38 with Topiroxostat (b), Fenbendazole (e), and Oxibendazole (h). Carbon atoms are shown in deep cyan, sulfur in yellow, oxygen in red, nitrogen in slate, and hydrogen in white. In silico predicted a two-dimensional schematic representation of the binding mode of CD38 in complex with Topiroxostat (c), Fenbendazole (f), and Oxibendazole (i). The color coding for the different types of intermolecular interactions is provided at the bottom of the figures.

MD simulations validated the dynamic stability of all three complexes over 100 ns (Figures S2–S4). For each system, the RMSD relative to the initial docked conformation rapidly converged and remained stable, indicating thermodynamically favorable equilibrium. The radius of gyration (Rg) of CD38 showed no significant fluctuations, confirming that ligand binding preserved overall protein compactness. Critically, the centroid distance between each ligand and the CD38 active site remained constant throughout the simulation, demonstrating stable anchoring at the catalytic pocket without dissociation. Buried solvent-accessible surface area (Buried SASA) maintained a steady contact interface between CD38 and each ligand, reflecting persistent intermolecular interactions. Conformational superposition of ligands extracted at different time points showed minimal structural deviation, further supporting stable binding. RMSF analysis revealed that active-site residues responsible for ligand recognition exhibited low fluctuation levels, consistent with stable inhibitor binding.

MM-PBSA calculations and residue-based energy decomposition quantified the energetic contributions to CD38-ligand binding (Table S1). Across all three complexes, van der Waals (VDW) interactions served as the dominant driving force, with ΔEvdw values ranging from approximately −126.9 to −149.5 kJ/mol. Electrostatic (ΔEele) and nonpolar solvation (ΔEnonpol) interactions acted as secondary stabilizing forces, while polar solvation (ΔEpol) consistently opposed complex formation. The Gibbs binding free energy (ΔGbind) confirmed highly favorable binding for all three ligands: −40.5 kJ/mol for Topiroxostat, −81.4 kJ/mol for Fenbendazole, and −83.8 kJ/mol for Oxibendazole, correlating with SPR-measured affinity differences. Residue-based decomposition identified TRP-189 and TRP-125 as major energetic contributors across all complexes. VAL-185 additionally stabilized the CD38–Topiroxostat complex, while the two tryptophan residues were the primary drivers for Fenbendazole and Oxibendazole binding. Collectively, these analyses demonstrate that stable CD38 inhibition by small-molecule ligands is achieved through a conserved binding mode centered on the enzyme’s active site, with van der Waals forces as the primary driving force and a set of aromatic and hydrophobic residues forming the essential ligand-recognition motif.

3. Discussion

In this study, we successfully applied DEL technology to discover novel small-molecule CD38 inhibitors, demonstrating the platform’s utility for targeting multifunctional enzymes. The DEL screening identified eight hit compounds with diverse chemical scaffolds, including clinically approved agents (e.g., Fenbendazole, Azathioprine), highlighting the potential for repurposing existing drugs for CD38-related diseases. This repurposing strategy offers advantages such as known safety profiles and reduced development timelines [24], which can accelerate the translation of lead compounds into clinical applications.

Fenbendazole emerged as a promising lead compound due to its moderate binding affinity, dual inhibitory activity against CD38 hydrolase and cyclase functions, and well-characterized pharmacokinetics. SAR analysis of Fenbendazole analogues revealed critical structural determinants of inhibitory potency: rigid aromatic substituents at the R1 position enhance binding via hydrophobic interactions with CD38 active-site residues, while the methoxycarbamide group at R2 is essential for hydrogen bonding. Flubendazole, a Fenbendazole analogue with a difluorophenyl R1 group, exhibited the highest potency, with IC₅₀ values in the low micromolar range. These findings provide a solid foundation for further optimization, such as modifying the R1 substituent to improve binding affinity or adding polar groups to enhance solubility, which could lead to the development of more effective CD38 inhibitors.

The dual inhibition of CD38 hydrolase and cyclase activities by our lead compounds is particularly notable. CD38’s hydrolase activity is the major contributor to NAD⁺ depletion in disease states, while its cyclase activity regulates calcium signaling via cADPR [2,7]. Inhibiting both activities may offer synergistic therapeutic benefits, especially in diseases like MM and age-related metabolic decline, where both NAD⁺ depletion and dysregulated calcium signaling play pathogenic roles [6,8]. This dual mechanism of action distinguishes our lead compounds from some existing CD38 inhibitors and may provide a competitive advantage in treating CD38-associated diseases.

Despite the significant progress made in this study, several limitations should be noted: (1) The DEL screening focused on extracellular CD38, and future studies should evaluate whether hit compounds can penetrate cells to target intracellular CD38 pools, as intracellular CD38 may also play a role in disease pathogenesis; (2) The inhibitory potency of lead compounds is in the micromolar range, and further optimization is needed to achieve nanomolar potency, which is typically required for effective therapeutic agents; and (3) In vivo validation is required to assess efficacy in CD38-dependent disease models (e.g., MM xenografts, aged mice) as well as evaluate the pharmacokinetic properties and potential off-target effects.

To address these limitations and advance the development of CD38 inhibitors, future directions include: (1) Structure-based optimization of Fenbendazole analogues using molecular docking and X-ray crystallography to improve binding affinity and selectivity; (2) Evaluation of lead compounds in cellular models of MM and age-related metabolic decline to confirm NAD⁺ restoration and functional improvements; (3) Investigation of compound selectivity against CD38 homologs (e.g., CD73, CD157) to minimize off-target effects [12], which is crucial for reducing potential side effects in clinical use; (4) In vivo pharmacokinetic and efficacy studies to assess therapeutic potential, including dose–response relationships, bioavailability, and tissue distribution.

In conclusion, our study established DEL technology as an efficient platform for CD38 inhibitor discovery and identified Fenbendazole analogues as promising lead compounds. These findings not only advance our understanding of CD38 inhibition but also provide a clear path toward developing novel therapeutics for CD38-associated diseases such as multiple myeloma and age-related metabolic decline [6,63,64]. With further optimization and validation, these lead compounds have the potential to make a significant impact on the treatment of these challenging diseases.

4. Materials and Methods

4.1. Chemicals and Materials

The compounds used in this study were obtained from commercial sources as follows: Alvimopan (MedChemExpress, Monmouth Junction, NJ, USA, #HY-76657A), Topiroxostat (MedChemExpress, USA, #HY-14874), Oridonin (TargetMol, Shanghai, China #T2790), Azathioprine (TargetMol, China, #T1237), Fenbendazole (MedChemExpress, China, #HY-B0413), Hyperoside (MedChemExpress, USA, #HY-N0452), Diacerein (MedChemExpress, USA, #HY-N0283), and Sparfloxacin (MedChemExpress, USA, #HY-B0308). Fenbendazole analogues were also obtained from commercial sources as follows: Albendazole (TargetMol, China, #T6372), Oxibendazole (MedChemExpress, USA, #HY-B0299), Parbendazole (MedChemExpress, USA, #HY-115364), Bendamustine (Cayman, Tianjin, China, #23693), Mebendazole (MedChemExpress, USA, #HY-17595), Flubendazole (J&K, Beijin, China, #129961), Oxfendazole (Aladdin, Shanghai, China, # O279005), Nocodazole (Beyotime, Shanghai, China, #S1765), and Thiabendazole (Yeasen, Shanghai, China, #60261ES76). All compounds were used as received without further purification and all compounds were dissolved in DMSO to prepare 10 mM stock solutions for subsequent use.

Nicotinamide guanine dinucleotide sodium salt (NGD), 1, N⁶-ethenonicotinamide adenine dinucleotide (ε-NAD), bovine serum albumin (BSA), sucrose, Tris Base (Trizma base), MES, sodium chloride, and other biochemical reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA). Bio-Rad Protein Assay Dye Reagent Concentrate was obtained from Bio-Rad Laboratories (Hercules, CA, USA). Surface plasmon resonance (SPR) chips were purchased from Cytiva (Chicago, IL, USA). The 96-well white opaque plates (Microfluor 1 White flat-bottom plate) were from Thermo Fisher Scientific (Waltham, MA, USA).

4.2. Protein and Library

The soluble human CD38 protein was prepared as follows: A synthetic gene was constructed to encode a signal peptide (MGWSCIILFVATATGVHS, derived from the mouse Igh-VJ558 gene) fused in-frame with the extracellular domain (residues R45-I300) of human CD38 (UniProt accession: P28907). This gene fragment was cloned into a mammalian expression vector between the NheI and HindIII restriction enzyme sites, resulting in the final expression construct pFuse-IgG1-Fc1. The recombinant soluble CD38 protein was produced by the transient transfection of Expi-293F cells with pFuse-IgG1-Fc1. Subsequent purification was performed using a sequential workflow: initial affinity chromatography with a Ni-Excel column, dialysis for buffer exchange, anion-exchange chromatography on an SP Sepharose HP column, and final size-exclusion chromatography using a Superdex 200 column (Cytiva) to obtain highly pure protein.

The nDEL (natural product and drug-like DNA-encoded library) was assembled following a late-stage DNA annotation method described previously [65]. It brings together 400 natural products, pharmaceutically active small molecules, and roughly 10⁴ combinatorial chemicals, with an emphasis on covering diverse chemotypes—including flavonoids, heterocycles, and other pharmacologically relevant scaffolds. Within the natural product subset, 110 highly purified compounds originated from traditional Chinese medicine (TCM), exemplified by oridonin, celastrol, and luteolin. The original report did not specify whether these were obtained commercially or purified in-house. The library also incorporates FDA-approved drugs and clinical-stage candidates, which were sourced either from standard commercial suppliers or through material transfer agreements with collaborators—a practice consistent with common research norms. As is typical in such descriptions, individual vendors are not enumerated. In-house synthesized materials constitute an integral part of the library. These include a 10⁴-member combinatorial DEL prepared internally, together with custom derivatives of known drugs (e.g., olaparib derivatives F001, F002, F003, F006) that served as positive controls. Additionally, the bifunctional diazirine linkers (compounds 1–6) essential for late-stage DNA labeling of all library members were also produced in-house.

4.3. DEL Screening Protocol

The DEL screening was performed following a standard affinity selection protocol with minor modifications. Briefly: (1) Immobilization of CD38: Recombinant human CD38 protein was immobilized on HisSep Ni-NTA MagBeads (Yeasen, # 20561ES03) according to the manufacturer’s instructions. A control resin without CD38 immobilization was prepared in parallel to eliminate non-specific binding. (2) Incubation: The immobilized CD38 resin (100 μL, 1 mg/mL protein) was incubated with the DEL (100 pmol total compounds) in binding buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween-20) at 4 °C for 1 h with gentle shaking. (3) Washing: The resin was washed sequentially with binding buffer (10 column volumes), binding buffer containing 300 mM NaCl (5 column volumes), and binding buffer (5 column volumes) to remove non-binding or weakly binding DEL compounds. (4) Elution: Bound DEL compounds were eluted with 200 μL of elution buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 M guanidine hydrochloride) and collected. (5) DNA extraction: The eluted solution was treated with phenol-chloroform to extract the DNA barcodes, which were then precipitated with ethanol.

4.4. PCR Amplification and High-Throughput Sequencing

The extracted DNA barcodes were amplified using PCR with primers specific to the common sequences of the DEL barcodes. The PCR reaction mixture (50 μL) contained 2 μL of DNA template, 25 μL of 2× Taq PCR MasterMix (Takara, Kyoto, Japan), 1 μL of each primer (10 μM), and 21 μL of ddH₂O. The PCR program was: 95 °C for 5 min; 30 cycles of 95 °C for 30 s, 55 °C for 30 s, 72 °C for 30 s; and 72 °C for 10 min. The PCR products were purified using a DNA purification kit (Qiagen, Germantown, MD, USA) and subjected to high-throughput sequencing on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA) with paired-end 150 bp reads.

4.5. Bioinformatics Analysis

The sequencing data were processed using custom bioinformatics pipelines to ensure the accurate identification of hit compounds. Low-quality reads (Phred score < 20) and adapter sequences were removed using Trimmomatic. The remaining reads were aligned to the reference barcode database of the DEL to identify the corresponding compounds. The alignment was performed using BLASTN with an E-value cutoff of 1 × 10⁻⁵. The enrichment ratio of each compound was calculated as the ratio of its abundance in the CD38-bound fraction to that in the control resin fraction. Compounds with an enrichment ratio > 2 and a read count > 100 were considered potential hit compounds [65]. To further refine the hit list, we excluded compounds that were enriched in multiple control screens (against unrelated proteins) to reduce the number of false positives. Additionally, compounds with known promiscuous binding properties (e.g., pan-assay interference compounds, PAINS) were identified using the PAINS filter and excluded from further analysis.

4.6. CD38 Enzymatic Activity Assays (Hydrolase and Cyclase Activities)

CD38 enzymatic activities were measured using fluorescence-based assays with ε-NAD (for hydrolase activity) and NGD (for cyclase activity) as substrates, respectively, with modifications based on previously described protocols. For assays using recombinant CD38 (including CD38-Fc fusion protein), the enzyme mix was prepared as follows: 10 ng/μL recombinant enzyme, 40 mg/mL BSA, and sucrose buffer, with a total volume of 50 μL per well. BSA was included in the enzyme mix to stabilize the enzyme and prevent non-specific adsorption to the plate. Test compounds (inhibitors or hit compounds) were diluted to 4× the desired final concentration in sucrose buffer (50 μL per well). For antibody-based inhibition, anti-CD38 antibodies were pre-incubated with recombinant CD38 (5 nM) at room temperature for 15 min at final concentrations of 200, 20, 2, and 0.2 nM (equal to 30, 3, 0.3, and 0.03 μg/mL). A blank control (50 μL sucrose buffer without test compounds) and a positive control (CD38 inhibitor 78c, 50 nM final concentration, or Isatuximab, 1 μg/mL) were included in each assay to validate the assay performance.

Assays were performed in 96-well white opaque plates with a final volume of 200 μL per well, in duplicates or triplicates to ensure reproducibility. Hydrolase Activity Assay: Reaction mix contained 50 μM ε-NAD (final concentration) in sucrose buffer. A total of 100 μL of normalized cell/tissue lysate, or 50 μL recombinant enzyme mix + 50 μL test compound, was added to each well. For blank wells, sucrose buffer was added instead of the sample/enzyme mix. Cyclase Activity Assay: Reaction mix contained 80 μM NGD (final concentration) in 20 mM Tris-HCl (pH 7.0). The setup was identical to the hydrolase assay.

The plate reader was configured to measure fluorescence at an excitation wavelength of 300 nm and emission wavelength of 410 nm. The analysis was set to kinetics mode, with readings every 30 seconds for at least 1 h, and a 5 second shake before reading to ensure uniform mixing. Then, 100 μL of the respective reaction mix was quickly added to all wells using a repeat pipette, the plate was loaded onto the reader tray, and measurements were initiated immediately.

Fluorescence readings from wells without CD38 (blank control) were subtracted from all sample readings to correct for background. Background-corrected fluorescence values (Relative Fluorescence Units, RFU) were plotted against reaction time. The initial reaction rate was calculated from the linear portion of the curve (typically the first 10–20 min) and used to determine the enzymatic activity. The percentage of enzymatic activity was calculated relative to the no-inhibitor control (DMSO-treated) using the formula: % Activity = (RFU of sample/mean RFU of no-inhibitor control) × 100. For dose–response analysis, IC₅₀ values were calculated using GraphPad Prism 9 software with non-linear regression analysis (log(inhibitor) vs. response—Variable slope). Statistical comparisons between groups were performed using the Student’s t-test, with a p-value < 0.05 considered statistically significant.

4.7. Binding Affinity Determination via SPR

Surface plasmon resonance (SPR) was used to measure the binding affinity of the active hit compounds and Fenbendazole analogues to CD38. Recombinant CD38 protein was immobilized on a CM5 sensor chip (Cytiva) via amine coupling. Briefly, the sensor chip was activated with a 1:1 mixture of 0.2 M N-ethyl-N’-(3-dimethylaminopropyl) carbodiimide (EDC) and 0.05 M N-hydroxysuccinimide (NHS) for 7 min at a flow rate of 10 μL/min. Recombinant CD38 protein was diluted to 50 μg/mL in 10 mM sodium acetate buffer (pH 4.5) and injected over the activated chip surface at a flow rate of 10 μL/min for 10 min to achieve an immobilization level of approximately 8000 resonance units (RU). The remaining active groups on the chip surface were blocked with 1 M ethanolamine hydrochloride (pH 8.5) for 7 min. A reference flow cell was prepared by following the same activation and blocking steps without immobilizing CD38 to correct for non-specific binding and bulk refractive index changes.

The hit compounds and analogues were serially diluted in running buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 0.05% Tween-20 and 5% DMSO) to concentrations ranging from 3.125 μM to 500 μM (or as indicated in specific experiments). A running buffer containing 5% DMSO (0 μM ligand) was used as the solvent control and subjected to the identical injection and detection procedures as the compound samples to eliminate potential non-specific signals induced by DMSO. All samples and the solvent control were injected over the chip surface at a constant flow rate of 30 μL/min. The association phase was monitored for 60 s, and the dissociation phase was monitored for 60 s. The chip surface was regenerated with 10 mM glycine-HCl (pH 2.5) for 30 s between consecutive injections to completely remove the bound compounds and restore the chip surface to its original state.

For sensorgram data processing, a three-step correction strategy was applied to ensure signal accuracy: first, the reference flow cell signal (blank chip surface without CD38 immobilization) was subtracted to eliminate non-specific binding to the chip matrix; second, the blank injection signal (running buffer without compound) was subtracted to remove background buffer-related signals; third, the 5% DMSO solvent control signal was subtracted to correct for solvent-induced baseline drift and non-specific interactions caused by DMSO. The fully corrected sensorgrams were fitted to a 1:1 Langmuir binding model using BIAevaluation software 3.2 (Cytica, Fremont, CA, USA). The reliability of affinity fitting for weak interactions (KD in the 10⁻⁴–10⁻⁵ M range) was verified using two criteria: (1) the goodness-of-fit was evaluated based on the correlation coefficient (R²) between the fitted curve and experimental data, with a threshold of R² > 0.95 for all accepted fitting results; (2) the residuals of the fitted curves exhibited random distribution around zero with no systematic deviation, confirming the appropriateness of the fitted model. Each binding experiment, including compound samples and solvent controls, was independently performed in technical duplicate, and the mean dissociation constant (K_D) value was reported for each compound.

4.8. Docking and Molecular Dynamics Simulation

Molecular dynamics simulations were performed using GROMACS 2022.2 [66]. The Amber14SB force field was assigned to the protein, and the TIP3P water model was used for the solvent phase. For the small-molecule ligand, AM1-BCC partial charges were generated via Antechamber with GAFF2 atom types assigned, and the corresponding GROMACS-compatible topology files were converted using ACPYPE. Ion parameters were consistent with the TIP3P model (Joung–Cheatham parameters). The protein–ligand complex was solvated in a truncated dodecahedral box with a minimum distance of 1.2 nm from the protein surface to the box boundary, and neutralized with Na⁺/Cl⁻ ions to a final concentration of 0.15 M.

Prior to production runs, the system was energy-minimized using the steepest descent algorithm until the maximum force was less than 1000 kJ·mol⁻¹·nm⁻¹. Subsequent equilibration was conducted in two sequential phases at 298 K: 200 ps of NVT (constant number of particles, volume, and temperature) followed by 200 ps of NPT (constant number of particles, pressure, and temperature) equilibration. Production MD simulations were run for 100 ns under NPT conditions with a 2 fs time step and Verlet cutoff scheme. Electrostatic interactions were calculated using the Particle Mesh Ewald (PME) method, with a cutoff radius of 1.2 nm for both Coulomb and van der Waals interactions. All hydrogen bonds were constrained using the LINCS algorithm. The system temperature was maintained at 298 K with the Nose–Hoover thermostat, and the pressure was kept at 1 bar using the Parrinello–Rahman barostat. Trajectory frames were saved every 10 ps for subsequent analysis.

Structural analysis and interaction characterization of the complex were performed using built-in GROMACS utilities 2024.6, VMD 1.9.4a57, and PyMOL 3.1.6.1 [67]. The Molecular Mechanics Poisson–Boltzmann Surface Area (MM-PBSA) method implemented in gmx_MMPBSA was used to calculate the binding free energy of the protein–ligand complex and decompose the energy contributions from individual amino acid residues.

5. Conclusions

Building upon our earlier work with DEL technology [68], this study further demonstrates its utility for the rapid discovery of novel CD38 inhibitors, establishing it as a highly efficient and cost-effective alternative to traditional high-throughput screening. By screening a diverse DNA-encoded library against recombinant human CD38, we identified eight hit compounds with distinct chemical scaffolds including clinically approved agents such as Fenbendazole. Through a comprehensive set of biochemical and biophysical assays, we confirmed that these hit compounds directly bind to CD38 and inhibit both its hydrolase and cyclase activities, highlighting their potential as dual-targeting therapeutics.

Fenbendazole, in particular, emerged as a promising lead compound due to its well-characterized safety profile and potent dual inhibitory activity. SAR analysis of Fenbendazole analogues further revealed critical structural features that govern CD38 binding affinity and inhibitory potency, with rigid aromatic substituents at the R1 position and the methoxycarbamide group at the R2 position being key determinants. Flubendazole, a Fenbendazole analogue with a difluorophenyl R1 group, exhibited the highest inhibitory potency, making it an ideal candidate for further optimization.

Molecular docking simulations provided valuable insights into the binding mode of lead compounds, confirming their interaction with key active-site residues of CD38 and explaining the observed SAR trends. These structural insights, combined with the experimental data, lay the groundwork for rational drug design and the further optimization of lead compounds to improve their binding affinity, selectivity, and pharmacokinetic properties.

While this study represents a significant step forward in CD38 inhibitor discovery, future work is needed to address the identified limitations, including evaluating the cell permeability of lead compounds, optimizing their inhibitory potency to the nanomolar range, and validating their efficacy in in vivo disease models. With continued development, these lead compounds have the potential to be translated into novel therapeutics for the treatment of CD38-associated diseases such as multiple myeloma, age-related metabolic decline, and neurodegenerative diseases, ultimately improving patient outcomes.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/molecules31050864/s1, Figure S1: SPR binding analysis of 78c to human CD38 extracellular protein; Figure S2: Molecular dynamics simulation analysis of the Topiroxostat-CD38 complex; Figure S3: Molecular dynamics simulation analysis of the Fenbendazole-CD38 complex; Figure S4: Molecular dynamics simulation analysis of the Oxibendazole-CD38 complex; Table S1: MM-PBSA binding energy components of the CD38-ligand complex (kJ mol⁻¹).

Author Contributions

Conceptualization, L.Y.; methodology, X.S. and Z.L.; software, W.M.; validation, G.Y. and L.Y.; formal analysis, L.Y.; investigation, X.S. and Z.L.; data curation, G.Y. and L.Y.; writing—original draft preparation, L.Y.; writing—review and editing, G.Y. and L.Y.; visualization, L.Y.; supervision, L.Y.; project administration, L.Y.; funding acquisition, G.Y. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The original contributions presented in this study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This work was supported by the Ministry of Science and Technology of China Fund (Grant No. 2021YFA1100800), the Jiangsu Basic Research Center for Synthetic Biology (Grant No. BK20233003), and the Shanghai Frontiers Science Center for Biomacromolecules and Precision Medicine.