Targeting undruggable carbohydrate recognition sites through focused fragment library design

Carbohydrate-binding proteins (lectins) mediate essential cell–cell and host–pathogen recognition events and are attractive therapeutic targets. Yet, their highly hydrophilic carbohydrate-recognition sites (CRDs) typically yield weak, polar interactions, making the discovery of drug-like, non-carbohydrate inhibitors challenging. Metal-dependent lectins employ Ca2+ ions to coordinate carbohydrate ligands, offering an opportunity to leverage metal coordination in ligand design. This study asks whether focused fragment libraries enriched in metal-binding pharmacophores (MBPs) can identify drug-like fragments that target Ca2+-dependent lectin CRDs and whether such fragments can be tuned for selectivity among diverse lectins (DC-SIGN, Langerin, LecA, LecB). The purpose is to establish MBPs (e.g., hydroxamates, malonates) as starting points for fragment-based drug discovery (FBDD) against lectin CRDs previously considered difficult to drug and to assess their biochemical, structural, and cellular activities.

Previous work highlights the therapeutic potential of targeting glycan–binding proteins (GBPs) in infection, immunity, and cancer, yet glycomimetics often suffer from poor pharmacokinetics and multivalency requirements. Prior fragment and computational efforts suggested druggable features in certain C-type lectins and identified catechols that coordinate Ca2+ in LecA and Langerin. Drug-like non-carbohydrate scaffolds and allosteric sites have also been explored for DC-SIGN and Langerin. Metal-ligand interactions are widely exploited in metalloenzymes (e.g., hydroxamates), but their application to non-enzymatic metal-binding proteins like lectins has not been demonstrated. These observations motivate a target-oriented MBP-enriched FBDD strategy for Ca2+-dependent lectins.

Design and libraries: Four fragment sources were used: computational virtual screening; a 3F Fsp3-rich 19F fragment library (115 fragments); a general fragment library (~1000 fragments, including 650 screened non-fluorinated fragments); and a Metal-Binding Pharmacophore (MBP) library (142 commercially available fragments from a chelator set; 98 fluorinated and 9 non-fluorinated retained after QC), featuring chelating motifs (picolinic acids, pyrimidines, hydroxypyrones, hydroxypyridinones, sulfonamides, β-diketones, salicylic and hydroxamic acids). Screening strategy: Fluorinated MBPs were mixed (30–32 per mix, 100 µM each) and screened by 19F NMR and 19F T2-filtered NMR in the presence/absence of proteins and CaCl2/EDTA to identify Ca2+-dependent binders; orthosteric competition was probed with natural carbohydrates (MeGal for LecA, MeFuc for LecB/BambL, D-mannose for DC-SIGN/Langerin). Non-fluorinated fragments were screened by 1H-15N HSQC/TROSY NMR against 15N-labeled targets. Targets: LecA and LecB from P. aeruginosa; mammalian C-type lectins DC-SIGN (CRD and ECD constructs) and Langerin (CRD and ECD); BambL as Ca2+-independent control. Biophysical validation and SAR: Protein-observed NMR (1H-15N HSQC/TROSY) to map CSPs; protein-observed 19F NMR (PrOF) using 5-fluorotryptophan-labeled LecA to monitor orthosteric W42; reporter-based 19F T2-filtered NMR using fluorinated “spy” analogs (hydroxamate 5; malonate 61) to rank analogs and determine Ca2+-dependency and competition; surface plasmon resonance (SPR) for LecA (limited ranking due to weak affinity); competitive fluorescence polarization (FP) for LecA inhibition vs galactose Cy5 probe; X-ray crystallography of LecA–hydroxamate complex; molecular docking (MOE) to rationalize binding modes in LecA, LecB, DC-SIGN; cell-based assay (cellFy) by flow cytometry to assess inhibition of FITC-dextran binding on DC-SIGN+ and Langerin+ Raji cells, and cytotoxicity. Data analysis: CSP thresholds ~0.01 ppm; hit confidence defined by 19F chemical shift and intensity changes; KD/Ka from one-site binding fits; ligand efficiency (LE) computed at experimental temperatures (298–310 K).

Druggability and hit rates: The MBP library substantially increased total hit rates for Ca2+-binding lectins (LecA, DC-SIGN, LecB) to 37–50%, outperforming general and other libraries; BambL (Ca2+-independent) showed distinct behavior. Hydroxamates for LecA: Hydroxamic acid scaffold 1 yielded multiple active analogs; 49 analogs assessed by TROSY NMR; 18 prioritized. Marketed metalloproteinase inhibitors (47, 49, 50) did not bind LecA, indicating steric/coordination mismatches. Competitive 19F NMR with spy 5 demonstrated that LecA–hydroxamate interaction is Ca2+-dependent and relatively selective versus LecB, DC-SIGN, and Langerin (the latter two showed Ca2+-independent weak interactions). FP assay showed LecA inhibition at 10 mM: compound 35 (26 ± 1%), 20 (33%), 36 (35 ± 3%), 6 (21 ± 1%); MeGal and GalNAc served as controls. PrOF NMR confirmed orthosteric binding via W42 perturbation; KD for 35 was 4.6 ± 0.9 mM; representative Ka and LE (Table 1): 1: Ka 7.2 ± 1.4 mM, LE 0.25; 6: 4.4 ± 0.6 mM, LE 0.30; 20: 4.5 ± 0.2 mM, LE 0.26; 35: 4.6 ± 0.9 mM, LE 0.26; 36: 3.6 ± 2.2 mM, LE 0.25. X-ray crystal structure (PDB 7FJH) of LecA–35 at 1.8 Å showed bidentate Ca2+ chelation by the hydroxamate, H-bonds to N107 and D100 (mimicking galactose OH3/OH4), and water-mediated network mimicking OH6; CH–π with P38. Malonates for one/two/three Ca2+ lectins: Malonate 58 and analogs bound LecA, LecB, and DC-SIGN orthosteric sites in a Ca2+-dependent manner; 19F spy 61 binding was displaced by natural ligands in LecA, LecB, DC-SIGN but not in Langerin/BambL (suggesting secondary-site binding there). LecA (one Ca2+): PrOF NMR indicated orthosteric binding (W42 CSPs); analog 61 produced strong effects. LecB (two Ca2+): 15N TROSY showed residues perturbed similarly to MeFuc in a Ca2+-dependent way; Ka values were low mM with flat SAR (e.g., 58: Ka 1.2 ± 0.4 mM; 64 reduced affinity 2.6 ± 0.6 mM), suggesting an electronegative group role (predicted CF2–T98 interaction). DC-SIGN (three Ca2+): 15N HSQC mapping with 58 perturbed the EPN-containing long loop (N344, N365) and D367 near Ca2+ site 1; CSPs resembled D-mannose; Ka for 58 was 1.2 ± 0.5 mM; analogs 58, 62, 69 interchangeable; electronegative ring substituents favored (predicted F313 interaction with CF2); methyl in 59 tolerated, offering a growth vector for DC-SIGN selectivity. Cellular activity: In cellFy, 58 dose-dependently inhibited FITC-dextran binding to DC-SIGN+ Raji cells with low cytotoxicity up to 10 mM and showed no effect on Langerin+ cells, mirroring selectivity seen in NMR. Selectivity trends: Hydroxamates showed selectivity for LecA among Ca2+-dependent lectins; malonate selectivity was tunable—61 favored LecA, whereas 58/62 favored LecB and DC-SIGN; malonates also exhibited a propensity for secondary sites in Langerin/BambL.

The study demonstrates that focusing fragment libraries on metal-binding pharmacophores can overcome the weak, polar nature of lectin CRDs by exploiting Ca2+ coordination. Hydroxamates, widely used against metalloenzymes, are shown here for the first time to target a non-enzymatic metal-binding protein (LecA), validating Ca2+-mediated anchoring in lectin CRDs, with structural confirmation and measurable ligand efficiencies. Although affinities are modest at the fragment stage, clear orthosteric engagement and emerging selectivity over other lectins underscore their utility as starting points, potentially translating to anti-biofilm strategies against Pseudomonas aeruginosa. Malonates provide a versatile scaffold able to coordinate one, two, or three Ca2+ ions across LecA, LecB, and DC-SIGN, with tunable selectivity via scaffold modifications and electronegative substituents enabling secondary interactions (e.g., CF2 with F313 in DC-SIGN or T98 in LecB). The MBP-enriched library markedly improved hit rates on Ca2+-dependent lectins, supporting a target-oriented MBP-FBDD approach. Cellular assays confirmed that malonate 58 can specifically inhibit DC-SIGN-mediated carbohydrate interactions in a cellular context with low cytotoxicity, aligning with biophysical observations.

Metal-binding pharmacophores are effective fragment scaffolds for targeting Ca2+-dependent lectin carbohydrate-recognition sites previously considered difficult to drug. Two chemotypes emerged: hydroxamates that selectively target LecA with crystallographically verified Ca2+ chelation and malonates that engage one to three Ca2+ sites across LecA, LecB, and DC-SIGN with tunable selectivity and cellular activity against DC-SIGN. The MBP-focused fragment library significantly increased hit rates, establishing a foundation for fragment growth and optimization toward drug-like lectin inhibitors. Future work should enhance potency through structure-guided growth from validated vectors (e.g., hydroxamate and malonate ring positions), exploit secondary interactions (e.g., CF2 contacts), resolve additional co-crystal structures (especially with LecB and DC-SIGN), and evaluate efficacy in relevant infection and immunology models.

Fragment affinities are in the mM range, limiting direct translation without optimization. SPR could not rank hydroxamates due to weak binding. Structure–activity relationships were often flat, particularly for LecB malonates. Protein backbone assignment for 15N LecB was incomplete, and co-crystallization for LecB complexes is ongoing, constraining detailed mechanistic interpretation. Some malonates showed binding to secondary sites in Langerin and BambL, indicating potential off-target or allosteric engagement. Cellular validation was limited to Raji cell lines with FITC-dextran readouts; no in vivo efficacy was assessed.