Synthesis and direct assay of large macrocycle diversities by combinatorial late-stage modification at picomole scale

Abstract

Macrocycles have excellent potential as therapeutics due to their ability to bind challenging targets. However, generating macrocycles against new targets is hindered by a lack of large macrocycle libraries for high-throughput screening. To overcome this, we herein established a combinatorial approach by tethering a myriad of chemical fragments to peripheral groups of structurally diverse macrocyclic scaffolds in a combinatorial fashion, all at a picomole scale in nanoliter volumes using acoustic droplet ejection technology. In a proof-of-concept, we generate a target-tailored library of 19,968 macrocycles by conjugating 104 carboxylic-acid fragments to 192 macrocyclic scaffolds. The high reaction efficiency and small number of side products of the acylation reactions allowed direct assay without purification and thus a large throughput. In screens, we identify nanomolar inhibitors against thrombin (K_i = 44 ± 1 nM) and the MDM2:p53 protein-protein interaction (K_d MDM2 = 43 ± 18 nM). The increased efficiency of macrocycle synthesis and screening and general applicability of this approach unlocks possibilities for generating leads against any protein target.

Fig. 1. Diversification of macrocyclic scaffolds by combinatorially appending fragments to peripheral groups.

a General principle of the approach. b Image of an 80 nL droplet transferred by ADE, shown in a 96-well plate and next to a 4 μL droplet for scale. The droplets contain fluorescein for visualization by UV light. Addition of target and assay reagent to 80 nL macrocycle reactions dilutes the organic solvent to 0.4% which is compatible with bioassays. c Model macrocycle scaffold 1 containing a peripheral primary amine (blue) that is modified by acylation. d Reaction of model macrocycle 1 with indicated acids 1–8, quantified by HPLC (absorbance and/or ion count). The first number indicates conversion at 4 µL volume via pipetting with DIPEA. The second and third numbers indicate conversion at 80 nL via acoustic liquid transfer with DIPEA and DABCO, respectively. e Randomly selected non-peptide scaffolds containing less accessible amino groups (in blue).

Fig. 2. Acylation of amines in macrocyclic scaffolds and library generation.

a Cyclative disulfide release of side-chain-deprotected peptides. b Formats of cyclic peptide scaffolds. Amino groups are shown in blue. c Amino acids used for scaffold synthesis. Lower case letters indicate D-amino acids. d Yields of the 45 tryptophan-containing scaffolds as determined by absorption measurement. e Carboxylic acids 9–16 that were used with acids 1–3 and 8 to diversify the scaffolds shown in panel b. f Schematic of the protocol for macrocycle library synthesis by acoustic liquid transfer with indicated reaction conditions.

Fig. 3. Screening of macrocyclic compound library against thrombin, hit identification, and structural analysis.

a Heat map with thrombin inhibition indicated for each macrocycle. The amino acid sequences of the scaffolds are provided in the Source Data file. b Replicate reaction (black) compared to original screen (green) of all macrocycles containing acid 14. For each of the two independent experiments, the reactions were performed once and the thrombin inhibition was measured once. c Chemical structures and activities of the top three hits M1, M2, and M3. The mean values and SDs of three independent measurements are shown. d Chromatographic separation of the acylation reactions to generate M1 and M2 and analysis of the fractions for thrombin inhibition. e X-ray structure of M1 bound to human thrombin at 2.27 Å resolution (PDB 6Z48) (https://www.rcsb.org/structure/6Z48). The inhibitor is shown in the space-filling model with the scaffold in blue and the carboxylic acid in yellow. f Zoomed in structure of thrombin with the sub-sites indicated. The chlorothiophene group fills the S1 sub-site and forms an interaction with Tyr228.

Fig. 4. Screen against MDM2 and hit characterization.

a The 192 macrocycle scaffolds (Supplementary Fig. 9) were combinatorially acylated with 104 carboxylic acids. The products were screened for displacement of a p53-based fluorescent peptide from human MDM2. Final macrocycle concentrations were 10 µM. For each macrocycle, the MDM2 binding was measured once. The amino acid sequences of the scaffolds are provided in Source Data file. b Displacement of the fluorescent reporter peptide from MDM2 by HPLC-purified hit macrocycles M6–M8 measured by FP. Mean values and SDs from three independent displacement assays are shown. c Chemical structures of fluorescein-labeled M6–M8 that were measured for affinity to MDM2 in a direct binding assay by FP. Mean values and SDs from three independent measurements of the binding by FP are shown.

Fig. 5. Affinity optimization of an MDM2:p53 inhibitor.

a Scaffolds of Library 3 are based on M8 wherein the amino acids shown in blue colors are diversified. Amino acid building blocks are shown in the three frames. b Heatmaps of MDM2 binding to the 63 scaffolds that were combinatorially acylated with 15 carboxylic acids at a 30 pmol scale. Binding to MDM2 was measured by displacement assay of the fluorescent peptide probe by macrocycles at a concentration of 750 nM. c Screening of Library 4 based on the best nine scaffolds from the previous screens and 15 additional carboxylic acids. d Binding of fluorescein-labeled and HPLC-purified macrocycle M10 (F-M10) to MDM2 as measured by FP. Mean values and SDs of three independent measurements of the binding by FP are shown. e Binding of unlabeled macrocycles M8 and M10 to MDM2 as measured by SPR.