Mixture-Based Screening of Focused Combinatorial Libraries by NMR: Application to the Antiapoptotic Protein hMcl-1

We report on an innovative ligand discovery strategy based on protein NMR-based screening of a combinatorial library of ∼125,000 compounds that was arranged in 96 distinct mixtures. Using sensitive solution protein NMR spectroscopy and chemical perturbation-based screening followed by an iterative synthesis, deconvolutions, and optimization strategy, we demonstrate that the approach could be useful in the identification of initial binding molecules for difficult drug targets, such as those involved in protein–protein interactions. As an application, we will report novel agents targeting the Bcl-2 family protein hMcl-1. The approach is of general applicability and could be deployed as an effective screening strategy for de novo identification of ligands, particularly when tackling targets involved in protein–protein interactions.

page S3 Structures of the 96 sulfonyl chlorides present in P1 in the library. Table S2 page S12 Mol% of each amino acid used for the preparation of mixtures and mmol used, considering a 0.1 mmol scale. Table S3 page S14 Analogues of Compound (12): SAR in P2, and relative CSP in the 1D-1 H-aliphatic spectrum. Table S4 page S16 SAR of Compound (20), and relative CSP in the 1D-1 H-aliphatic spectrum. Table S5 page S22 Mass-spectrometry data of compounds. All the compounds were analyzed using an Agilent 6545 QTOF LC/MS instrument. Figure S1 page S24 Deconvolution approach used to identify P2 and P3. Figure S2 page S25 Experimental data of the second step of the deconvolution approach. Figure S3 page S26 Experimental data of the third step of the deconvolution approach: mixtures in P2. Figure S4 page S27 Experimental data of the deconvolution approach for the negative mixture D05. Figure S5 page S28 Experimental data of the addition of another amino acid in position 4 (P4) using the same deconvolution approach. Figure S6 page S29 Experimental data of the identification of D-Trp as the best amino acid in P4. Figure S7 page S30 Biophysical characterization of Compound (21) binding to hMcl-1.

Figure S8
page S31 ITC curves of hMcl-1 in the presence of Compound (59) and Compound (60). Figure S9 page S32 2D [ 1 H, 13 C] correlation spectrum of 20 μM 13 C-e-Met hMcl-1 in presence of different concentration of Compound (50). Figure S10 page S33 HPLC trace for Compound (21). Figure S11 page S34 HPLC trace for Compound (50) (peak 1). Figure S12 page S35 HPLC trace for Compound (51) (peak 2).   Table S2. Relative equivalents used of each amino acid in the preparation of mixtures. The exact amounts used in our 0.1 mmol scale is also reported.     Figure S1. Schematic representation of the deconvolution approach used for the HIT identification: one position at a time, P3, as in this case, or P2, is deconvoluted in order to find the best amino acid that fits that position. In the first step of the deconvolution approach, 4 submixtures are synthesized with the sulfonamide in P1, Ala in P2 and a mixture of nine amino acids in P3 (S1 A). Those four sub-mixtures are analyzed by NMR and the one that creates the bigger CSP in the 1D-1 H-aliphatic spectrum is further deconvoluted through the synthesis of individual compounds (S1 B). The individual compounds are then tested by NMR to identify the best amino acid that fits P3; the same deconvolution approach is then repeated to study P2, fixing the sulfonamide in P1 and the amino acid found previously in P3.

N/A +++
S25 Figure S2. Experimental data of the second step of the deconvolution approach. A) Structures of the compounds synthesized with E07 sulfonamide in P1, Ala in P2 and individual amino acids of the sub-mixture 4 in P3. B) Analysis by NMR of 20M hMcl-1 alone (blue) and in presence of each individual compound (250M). Compound (7) is the compound that presents the bigger CSP in the 1D-1 H-aliphatic region of hMcl-1.
S26 Figure S3. Experimental data of the third step of the deconvolution approach: study of P2. A) Structures of the four sub-mixtures synthesized with E07 sulfonamide in P1, sub-mixtures of amino acids in P2 and Cha in P3. B) Analysis of the four sub-mixtures by NMR: representation of the CSP that they cause in the 1D-1 H-aliphatic region of the hMcl-1 spectrum.
S27 Figure S4. A) Aliphatic region of the 1D 1 H NMR spectra of hMcl-1 (20 M) recorded in the absence (blue) and presence (red) of mixture D05 (2 mM). B) 1D 1 H NMR spectra of the sub-mixtures synthesized with D05 sulfonamide in P1, Ala in P2 and four smaller submixtures in P3. C) 1D 1 H NMR spectra of the sub-mixtures synthesized with D05 sulfonamide in P1, four smaller submixtures in P2 and Ala in P3. Compound (21) is the compound that presents the bigger CSP in the 1D-1 H-aliphatic region of hMcl-1. Chemical shift differences of Met 231 resonances between the free vs Compound (50) bound form provide an estimated upper limit for the off rate for the complex of koff < 519 s -1 , that assuming a diffusion limited on the rate of 10 9 M -1 s -1 , would correspond to a low micromolar dissociation constant Kd.