In vitro selection of a DNA-templated small-molecule library reveals a class of macrocyclic kinase inhibitors

Abstract

DNA-templated organic synthesis enables the translation of DNA sequences into synthetic small-molecule libraries suitable for in vitro selection. Previously, we described the DNA-templated multistep synthesis of a 13,824-membered small-molecule macrocycle library. Here, we report the discovery of small molecules that modulate the activity of kinase enzymes through the in vitro selection of this DNA-templated small-molecule macrocycle library against 36 biomedically relevant protein targets. DNA encoding selection survivors was amplified by PCR and identified by ultra-high-throughput DNA sequencing. Macrocycles corresponding to DNA sequences enriched upon selection against several protein kinases were synthesized on a multimilligram scale. In vitro assays revealed that these macrocycles inhibit (or activate) the kinases against which they were selected with IC(50) values as low as 680 nM. We characterized in depth a family of macrocycles enriched upon selection against Src kinase, and showed that inhibition was highly dependent on the identity of macrocycle building blocks as well as on backbone conformation. Two macrocycles in this family exhibited unusually strong Src inhibition selectivity even among kinases closely related to Src. One macrocycle was found to activate, rather than inhibit, its target kinase, VEGFR2. Taken together, these results establish the use of DNA-templated synthesis and in vitro selection to discover small molecules that modulate enzyme activities, and also reveal a new scaffold for selective ATP-competitive kinase inhibition.

Figure 1

The 13 824-membered DNA-templated small-molecule macrocycle library. (a) Scheme for the multistep DNA-templated synthesis of the macrocycle library.(24) (b) Amino-acid building blocks used in the library synthesis. Step one building blocks (A1−A12) are shown in red, step two building blocks (B1−B12) are shown in blue, step three building blocks (C1−C12) are shown in green, and the macrocycle scaffolds (D1−D8) are shown in black.

Figure 2

In vitro selection of the DNA-templated macrocycle library against 36 protein targets. Following affinity selection of the library against an immobilized GST-fused target protein, library members possessing target affinity were eluted and their attached DNA templates were amplified by PCR using DNA primers containing a barcode sequence that uniquely identifies the protein target. Barcoded PCR amplicons from 36 target protein-binding selections, one control selection for binding immobilized GST, and eight aliquots of unselected library were pooled and submitted as one sample for ultra-high-throughput DNA sequencing.

Figure 3

Analysis of high-throughput sequencing results. Plot of enrichment factor vs sequence abundance for library members after selection for binding to Src kinase (enrichment vs sequence abundance plots for all other selections performed are shown in Figure S9). The enrichment factor for each library member in each selection (∼497 000 total possible combinations of library member and target) were calculated as (postselection abundance)/(preselection abundance) using a PERL script. Each blue or red dot represents the DNA sequence corresponding to a single library member. The structures of macrocycles 1−9 are shown in Figure 4. Enrichment factors for low-abundance library members vary widely due to statistical undersampling. Only enrichment factors that were substantially higher than background values were considered potential positives (shown in red).

Figure 4

Chemical structures of macrocycles exhibiting significant enrichment above background and common structural motifs after selection against Src kinase. Macrocycle numbering corresponds to that used in Figure 3.

Figure 5

Plots of enrichment factor vs sequence abundance for library members after selection for binding to Akt3, MAPKAPK2, Pim1, and VEGFR2. Macrocycles 10−19 exhibit significant enrichment above background.

Scheme 1. Milligram-Scale Synthesis of Macrocycle A11-B1-C5-D7

The route was adapted from Gartner et al.(20) Methyl hydrogen 2,3-O-isopropylidene-(L)-tartrate was synthesized as reported by Musich and Rapoport.(70)

Figure 6

Macrocycle cis-A11-B1-C5-D7 is an ATP-competitive inhibitor of Src kinase. (a) The apparent K_M of Src for ATP was measured in the presence of increasing concentrations of macrocycle cis-A11-B1-C5-D7. Kinase activity was measured as in Table 3. (b) The resulting relationship is linear in accord with the following equation for a classical competitive inhibitor: apparent K_M = K_M × (1 + [inhibitor]/K_i).

Figure 7

Selectivity of macrocycles cis-A11-B1-C5-D7 and trans-A10-B1-C5-D6 among Src-related kinases, including all nine Src-family kinases. Both macrocycles were assayed at 5 μM concentration against the indicated kinases. Assays of cis-A11-B1-C5-D7 against Abl and Hck were performed at 50 μM macrocycle concentration instead of 5 μM. Percent kinase inhibition values plotted are the average of two independent measurements. Assay data for Src kinase is from Figure S2 and Table 3. All non-Src assay points were measured by the Invitrogen Select Screen Profiling Service.

Figure 8

SAR analysis of Src-inhibiting macrocycles. (a) Single-alanine mutants of cis-A11-B1-C5-D7 were assayed against Src-family kinases Src and Fgr at 5 μM concentration as described previously. (b) Linear diacetylated peptides corresponding to macrocycles A11-B1-C5-D7 and A10-B1-C5-D6 were synthesized by Fmoc solid-phase peptide synthesis and assayed against Src kinase. (c) Macrocycle trans-Phe-B1-C5-D6 (replacing nitrophenylalanine with phenylalanine) was assayed against Src as described previously.