Discovery of Novel Lin28 Inhibitors to Suppress Cancer Cell Stemness

Abstract

Lin28 is a pluripotency factor that regulates cancer cell stem-like phenotypes to promote cancer development and therapy-resistant tumor progression. It acts through its cold shock domain and zinc knuckle domain (ZKD) to interact with the Let-7 pre-microRNA and block Let-7 biosynthesis. Chemical inhibition of Lin28 from interacting with Let-7 presents a therapeutic strategy for cancer therapy. Herein, we present the computer-aided development of small molecules by in silico screening 18 million compounds from the ZINC20 library, followed by the biological validation of 163 predicted compounds to confirm 15 new Lin28 inhibitors. We report three lead compounds, Ln7, Ln15, and Ln115, that target the ZKD of both Lin28A and Lin28B isoforms and block Lin28 from binding Let-7. They restore Let-7 expression and suppress tumor oncogenes such as SOX2 in cancer cells and show strong inhibitory effects on cancer cell stem-like phenotypes. However, minimal impacts of these compounds were observed on Lin28-negative cells, confirming the on-target effects of these compounds. We conclude from this study the discovery of several new Lin28 inhibitors as promising candidate compounds that warrant further drug development into potential anticancer therapies.

Keywords: Let-7; Lin28 inhibitor; RNA binding inhibitor; cancer stem cell; zinc knuckle domain.

Figure 1

(A) The CSD (green) and ZKD (blue) domains interact with pre-Let-7 (white ribbon) in a bipartite manner (PDB: 5UDZ). The GGAG motif (red) and the surface of the proposed binding pocket (green) are marked. (B) A DeepDocking workflow is shown. First, for each molecule in the database molecular descriptors are calculated. These descriptors are then used to cluster all molecules. Second, molecules are sampled from clusters and docked into the target protein. Third, the derived docking scores and corresponding molecular descriptors are used to build machine learning models that predict the docking scores of undocked compounds. With each iteration, docking scores are predicted for a batch of molecules and the poorly scored molecules are removed. The number of iterations is user-defined for 5 rounds. The output is a list of molecules with predicted plausible docking scores. (C) The CADD pipeline is employed to predict 163 potential Lin28 ZKD inhibitors for biological testing. (D) Interaction between Lin28B ZKD and the two nucleotides (first guanine and the adenosine of the GGAG motif) from Let-7. The residues that form hydrogen bonds (in red dotted lines) are used as pharmacophore features and labeled by red boxes.

Figure 2

Establishment of the Fluorescence Polarization Assay and Initial compound screening. (A) Graphical representation of the FP assay principle. As the labeled probe floats freely in the solution, it rotates more rapidly, producing a low FP signal. When the labeled probe binds to proteins, it rotates slower and produces a high FP signal. (B) In FP assays, probe titration was performed by using 10 uM protein with 1 nM–100 uM FAM-labeled Let-7 probe (n = 9 replicates). Protein titration was performed by using 1 nM of FAM-labeled Let-7 probe with 1 nM–50 uM Lin28B ZKD protein (n = 9 replicates), and competition titration using 1 nM–50 µM non-labeled Let-7 probe with 10 uM of Lin28B ZKD protein with 1 nM FAM-labeled Let-7 probe (9 replicates). (C) FP assays were used to screen 163 compounds at the concentration of 20 uM using 1 nM of FAM-labeled Let-7 probe with 10 uM Lin28B ZKD protein.

Figure 3

Lin28 inhibitors block the interaction between Lin28B ZKD and Let-7. (A) EMSA used 10 nM of IR800DyeCWN-labeled let-7 probe (lane 1), 0.5 uM Lin28B ZKD protein with 10 nM of labeled Let-7 probe (lane 2), or Lin28B ZKD, labeled Let-7 probe and 100× non-labelled competitor probe (lane 3). EMSA used Lin28B ZKD, labeled Let-7 probe, and 100–1200 uM of indicated compounds (lanes 4–10). (B) FP assays were performed by using 10 uM of Lin28B ZKD and 1 nM of FAM-labeled Let-7 probe and 1 nM–1000 uM of compounds. (C) Bio-layer interferometry (BLI) assays were performed using 0–25 uM of Ln115 with Avi-tagged Lin28B ZKD protein. (D) Chemical structures of the three lead compounds were shown.

Figure 4

Lin28 inhibitors restore Let-7d expression and suppress SOX2 and HMGA2 genes. DUNE (top) and DUNE (KO) (bottom) cells were treated with 20 uM of the indicated compounds for 48 h. Real-time qPCR measured RNA levels of Let-7d, SOX2, and HMGA2. Three biological replicates were performed and all results are presented as mean +/− SD. p-value ≤ 0.05—(*); ≤0.01—(**); ≤0.001—(***).

Figure 5

Lin28 inhibitors block the Lin28A isoform. (A) EMSA used 10 nM of IR800DyeCWN-labeled Let-7 probe (lane 1), 0.5 uM Lin28A ZKD protein with 10 nM of labeled Let-7 probe (lane 2), or Lin28A ZKD, labeled Let-7 probe and 100× non-labeled competitor probe (lane 3). EMSA used Lin28A ZKD, labeled Let-7 probe, and 100–1200 uM of indicated compounds (lanes 4–10). (B) FP assays were performed by using 10 uM of Lin28A ZKD, 1 nM of FAM-labeled Let-7 probe, and 1 nM–200 uM of compounds. (C) Lin28A-positive IGROV1 cells were treated with 20 uM of the indicated compound for 48 h. Real-time qPCR measured RNA levels of Let-7d, SOX2, and HMGA2. Three biological replicates were performed and all results are presented as mean +/− SD. p-value ≤ 0.05—(*); ≤0.01—(**); ≤0.001—(***).

Figure 6

Lin28 inhibitors block the expression of NE and CSC biomarkers. DUNE cells were treated with 20 uM of the indicated compound for 48 h. Real-time qPCR measured RNA levels of neuroendocrine (A) and CSC (B) biomarkers. Three biological replicates were performed and all results are presented as mean +/− SD. p-value ≤ 0.05—(*); ≤0.01—(**); ≤0.001—(***).

Figure 7

Lin28 inhibitors block CSC phenotypes of cancer cells. (A) FACS analysis using CD44 and CD133 antibodies detected CSC cell populations from DUNE or DUNE (KO) cells treated with either vehicle or 20 uM of Ln7, Ln15, and Ln115 for 48 h. (B) DUNE, DUNE (KO), and IGROV1 cell proliferation were measured in the presence of 20 uM of Ln7, Ln15, and Ln115. (C) Colony formation assays (colonies >50 uM were counted) measured the suppressive effects of Ln7, Ln15, and Ln115 at the concentration of 20 uM. p-value ≤ 0.01—(**); ≤0.001—(***).