Flexibility-tuning of dual-display DNA-encoded chemical libraries facilitates cyclic peptide ligand discovery

Abstract

Cyclic peptides constitute an important drug modality since they offer significant advantages over small molecules and macromolecules. However, access to diverse chemical sets of cyclic peptides is difficult on a large library scale. DNA-encoded Chemical Libraries (DELs) provide a suitable tool to obtain large chemical diversity, but cyclic DELs made by standard DEL implementation cannot efficiently explore their conformational diversity. On the other hand, dual-display Encoded Self-Assembling Chemical (ESAC) Libraries can be used for modulating macrocycle flexibility since the two displayed peptides can be connected in an incremental fashion. In this work, we construct a 56 million dual-display ESAC library using a two-step cyclization strategy. We show that varying the level of conformational restraint is essential for the discovery of specific ligands for the three protein targets thrombin, human alkaline phosphatase and streptavidin.

Fig. 1. DP-DEL library architecture and synthesis.

A Schematic representation of DP-DEL synthesis. The building blocks (BB1-5) and their corresponding DNA codes are color-coded. The library is formed by the hybridization of the synthesized sub-libraries HP5 and HP3 and optional crosslinking. Fmoc-L-Cys(oNv)-OH was conjugated as a scaffold to both partially complementary 5’ NH₂-modified 38-mer HP5 and 3’ NH₂-modified 38-mer HP3, followed by Fmoc group removal (I). BB1 and BB3 were conjugated via amide bond formation onto the scaffolds of HP5 and HP3 strands, respectively, followed by Fmoc removal and HPLC purification of each conjugate. Enzymatic adaptor-mediated ligation was performed for the encoding of each building block, before pooling HP5 and HP3 conjugates, respectively (II). The HP5 strand was sequentially reacted with BB2 which were attached using amide bond formation, followed by enzymatic adaptor-mediated ligation, Fmoc deprotection and diazotransfer onto the free amino function (III). The HP3 strand was pre-encoded with BB4 codes, followed by amide bond formation of BB4 and piperidine treatment to remove DMT-MM adducts (IV). HP5- and HP3-conjugate pools were HPLC purified, then the two sub-libraries were mixed stoichiometrically, allowing the HP5 and HP3 strands to hybridize, resulting in version 1 of DP-DEL. The major part of the hybridized product was subjected to an interstrand copper-catalyzed Huisgen cycloaddition (CuAAC) reaction to link the azide functionality of BB2 with the alkyne functionality of BB4, resulting in covalently linked HP5-HP3 strands, forming a semi-closed peptide macrocycle (version 2 of DP-DEL) (V). This product was split again, subjecting the major part of the library version 2 to Cys(oNv) group removal, allowing the scaffolds of both sub-libraries to connect through either a disulfide or through the addition of 3 different bis-electrophiles, reacting with both cysteine scaffolds simultaneously (VI). This resulted in fully-closed peptide macrocycles between the two HP5/HP3 sub-libraries (version 3 of DP-DEL). A final step of encoding followed, where DP-DEL versions 1-3 could be encoded as building block 5 via hybridization of different codes 5 and a junction primer. Through T4 DNA polymerase and T4 ligase activity, the non-complementary encoding information of HP5 and HP3 sub-libraries + codes 5 could be transformed into a single continuous DNA strand. After this operation, all different versions of the DP-DEL were pooled to result in a finished master library. B Characteristics of building blocks used for the creation of DP-DEL. Building blocks 1-3 consist of a selection of natural and non-natural amino acids and dipeptides. BB4 consists of alkyne carboxylic acids and BB5/code 5 define versions 1-3 by varying the interaction between the two sub-libraries HP5 and HP3. C) Distribution of molecular weights for the fully cyclized library members (version 3 of library).

Fig. 2. High-throughput DNA sequencing fingerprints of naïve DP-DEL and control selections.

A High-throughput DNA sequencing (HTDS) results of unselected DP-DEL represented through a 3D plot displaying the distribution of the most prominent BBs 1-3 where each axis represents a BB position and each dot is a library member carrying a combination of said BBs, while the heat bar represents sequence counts. No cut-off was applied for unselected library. B, C Positive control concatenated triplicate HTDS against CAIX and NSP14 respectively. Spiked-in sulfonamide BB1-175 and BB3-176 showed preferential enrichment as binders for CAIX. Spiked-in S-adenosylhomocysteine BB1-119 and BB3-137 as a combination showed preferential enrichment as binders against NSP14. A cut-off of 20 counts was set for CAIX and NSP14. Selections were performed in triplicate and concatenated into a single dataset for visual display (for the selection results with the individual DP-DEL versions see Supplementary Information Fig. S28).

Fig. 3. DP-DEL thrombin selections.

A HTDS of concatenated triplicate selections against thrombin. Building blocks 1, 2 and 3 correspond to each respective axis. The heat bar represents the DNA sequence counts. The Plot is constrained to combinations from version 2 of DP-DEL possessing BB4-position 2, due to the attainment of highest enrichment values from that version of the library (see Supplementary Information, section 5.4.2.). Arrows indicate the three most enriched library members from the selection. Cut-off: 100 counts. B On-LNA experimental validation of highest enriched library members 1-3 displayed on 8-mer locked nucleic acid (LNA) heteroduplexes with a fluorescein label by fluorescence polarization (FP). Conjugates were tested in open form approximating version 1 of DP-DEL and clicked form approximates version 2 of DP-DEL. Affinity values are given as a mean of triplicate experiments (n = 3), with the exception of 1-o, which was measured as a duplicate (n = 2). C Off-DNA inhibitory measurements of selected compound 4 (R = Ahx-(H-Lys-NH₂)-5-carboxyfluorescein) against human alpha-thrombin, synthesized using a selection of linkers between the two cysteine scaffolds to test preferred peptide flexibility. L1-L5 allow the formation of variably sized macrocyclic peptides, while L0 yields a linear peptide. A heat bar shows the relation of IC₅₀ obtained and peptide steric constraint due to L0-L5 of binder. IC₅₀ values are obtained from experimental triplicates (n = 3) and given as a mean. Error bars indicate standard deviation of the replicates.

Fig. 4. DP-DEL streptavidin selections.

A DP-DEL HTDS of concatenated triplicate selections against streptavidin. Plot is constrained to combinations from version 2 of DP-DEL possessing BB4-position 2, due to the attainment of highest enrichment values from that version of the library (see supplementary section 5.4.2.). Arrows indicate the 2 highest enriched library members from the selection. Cut-off: 100 counts. B On-LNA experimental validation of highest enriched library members 5 & 6 displayed on 8-mer locked nucleic acid (LNA) heteroduplexes with a fluorescein label by fluorescence polarization (FP). Conjugates were tested in open (5-o, 6-o) form approximating version 1 of DP-DEL and clicked (5-c, 6-c) form approximates version 2 of DP-DEL. Affinity values are given as a mean of a duplicate or triplicate experiments. C Off-DNA affinity measurements of isolated streptavidin binders 7 & 8 (R2 = Ahx-(H-Lys-NH2)-5-carboxyfluorescein). Preferred peptide flexibility is examined through the synthesis of compound 7 using linkers L0-L5 and affinity of 8 is examined through synthesis with best performing L0. Selectivity of both 7-L0 & 8-L0 is tested against 3 different biotin-binding proteins, showing specificity towards streptavidin.

Fig. 5. DP-DEL PLAP selections.

A DP-DEL HTDS of concatenated triplicate selections against PLAP. Building blocks 1, 2 and 3 correspond to each respective axis. The heat bar represents DNA sequence counts. Plot is constrained to combinations from version 1 of DP-DEL, for both BB4 positions (see supplementary section 5.4.2.). Arrows indicate the 3 highest enriched library members from the selection. Cut-off: 30 counts. B On-LNA experimental validation of highest enriched library members 9-14 displayed on 8-mer locked nucleic acid (LNA) heteroduplexes with a fluorescein label by fluorescence polarization (FP). Conjugates were compared in their version 1 form with an azide group at the BB2 position or an NH₂ group. C Optimization of dual display partners for best performing binders 10 and 12. Two additional enriched BB3 + BB4 combinations as well as a negative control were tested to improve on binders 10 and 12. D Testing of different length maleimide linkers between cysteine scaffolds of both strands, to determine optimal spacing of best performing dual display partners 17. E Off-DNA resynthesis and inhibition experiments of optimized binder 22 as well as its single display variant 21 (R₂ = Ahx-(H-Lys-NH₂)-FITC). Affinity and IC₅₀ values from experimental triplicates (n = 3) are given as a mean. Error bars indicate standard deviation of the replicates.