Harnessing Connectivity in a Large-Scale Small-Molecule Sensitivity Dataset

Abstract

Identifying genetic alterations that prime a cancer cell to respond to a particular therapeutic agent can facilitate the development of precision cancer medicines. Cancer cell-line (CCL) profiling of small-molecule sensitivity has emerged as an unbiased method to assess the relationships between genetic or cellular features of CCLs and small-molecule response. Here, we developed annotated cluster multidimensional enrichment analysis to explore the associations between groups of small molecules and groups of CCLs in a new, quantitative sensitivity dataset. This analysis reveals insights into small-molecule mechanisms of action, and genomic features that associate with CCL response to small-molecule treatment. We are able to recapitulate known relationships between FDA-approved therapies and cancer dependencies and to uncover new relationships, including for KRAS-mutant cancers and neuroblastoma. To enable the cancer community to explore these data, and to generate novel hypotheses, we created an updated version of the Cancer Therapeutic Response Portal (CTRP v2).

Significance: We present the largest CCL sensitivity dataset yet available, and an analysis method integrating information from multiple CCLs and multiple small molecules to identify CCL response predictors robustly. We updated the CTRP to enable the cancer research community to leverage these data and analyses.

Figure 1

The Informer Set comprises 481 small-molecule compounds targeting a wide range of proteins involved in cell growth and survival. Sunburst visualization of validated protein targets of the small molecules in the Informer Set using protein family hierarchy from the Panther database (12). Approximately 115 compounds within the Informer Set do not have validated protein targets (nMoA). These compounds were included because they are known to affect a specific process or pathway, or to elicit gene-expression responses not seen in compounds having known mechanisms of action; however, they are not represented in this visualization.

Figure 2

ACME analysis identifies hotspots that link genetic features of CCLs to patterns of small-molecule response. A, clustered AUC matrix for 481 small molecules and 664 adherent, genomically characterized CCLs, and the corresponding dendrograms (blue, CCLs; black, small molecules). Gray represents AUC values that were not measured or did not pass QC metrics. B, visualization of ACME analysis, which requires enrichment of small-molecule annotation, CCL annotation, and sensitivity of CCLs to the small molecules. Depicted are data for the association of sensitivity of chronic myeloid leukemia cell lines to treatment with inhibitors of ABL1. Both compound and CCL clusters (red) and the area of sensitivity in the AUC matrix (black) are depicted. C, hotspot and corresponding dendrograms for MEK inhibitors and BRAF^V600 CCLs (left), including AUCs corresponding to the intersection of the two clusters (black box). The distributions of the AUCs for compound clusters, as well as the null distributions, can be visualized either with a fitted density curve (middle; bin number selected to scale with the number of CCLs), or empirical cumulative distribution function (CDF) plot (right). Each dendrogram segment is marked with the corresponding maximum height. The purity of both the compound cluster and the CCL cluster is 1. D, association of CCLs from the breast lineage with response to ERBB2 inhibitors. Visualization of the hotspot (black box) in the AUC matrix with the corresponding dendrograms (left), and the corresponding empirical CDF plot (right). The purity of both the compound cluster and the CCL cluster is 1. E, ERBB2 expression from CCLE for available adherent CCLs and the CCLs in the breast CCL-enriched cluster.

Figure 3

ACME analysis sheds light on small-molecule mechanism of action. A, dendrogram of the small-molecule Informer Set. Colored dots on the dendrogram denote the location of the enlarged dendrogram segments displayed to the right (color-coded to boxes). The top left inset (red box) is a “zoom in” of the cluster enriched for PI3K signaling, and the bottom left inset (blue box) contains one cluster enriched for MDM2 and one enriched for bromodomain inhibitors. The right inset (orange box) is the antimitotic cluster discussed in the text. For this inset, protein targets are colored by class: inhibitors of microtubule assembly (black), antimitotic kinase inhibitors with targets other than tubulin (red), and compounds with nominal protein targets unrelated to microtubule assembly or mitotic kinases (orange). B, recombinant tubulin polymerization assay. Every third data point is displayed. Each compound was run in duplicate (paclitaxel, 3 μmol/L; nocodazole, 3 μmol/L; KX2-391, 10 μmol/L; YK-4-279, 10 μmol/L; NVP-231, 10 μmol/L; LY2183240, 10 μmol/L). RFU, relative fluorescence units. C, microtubule regrowth assay in NCIH661 cells. Cells were cooled for 30 minutes on ice prior to compound treatment at time 0, and cells were either fixed directly or warmed to 37°C for 5 minutes and then fixed, followed by immunostaining for nucleus/DNA Hoechst stain (blue) and α-tubulin (green). All compounds were used at 500 nmol/L. D, BRD4 bromodomains 1 and 2 time-resolved FRET assay. E, Western blot of lysates from MM1S cells treated for 6 hours with compound at the indicated concentrations.

Figure 4

ACME analysis associates the sensitivity of neuroblastoma CCLs to IGF1R inhibitors to reveal the exquisite sensitivity of ALK-overexpressed neuroblastoma CCLs to dual IGF1R and ALK inhibition. A, the three enrichments for IGF1R inhibitors and neuroblastoma CCLs: row and column dendrograms and empirical cumulative distribution function (CDF) of the AUC distributions. The purity of the compound cluster is 0.8 and the confdence is 1. The purity of the CCL cluster is 0.55 and the confdence is 0.43. On the compound dendrogram segment, IGF1R inhibitors (black) and NVP-TAE684 (inhibiting ALK; red) are depicted. B, Western blot of lysates from NB1 cells that were serum-starved overnight, followed by 3-hour treatment with NVP-TAE684 at the indicated concentrations, and then 10-minute stimulation with IGF1. Experiments were repeated twice. These data confrm loss of phophorylated (p) IGF1R upon treatment with NVP-TAE684. C, confrmation of profling results for BMS-754807, crizotinib, and NVP-TAE684. The average of two replicates from two independent experiments is shown. D, AUC-AUC comparison for NVP-TAE684 and BMS-754807 with neuroblastoma CCLs (dark blue), CCLs with NPM-ALK rearrangement (light blue), and EML4-ALK rearrangements (orange) highlighted. E, sensitization of crizotinib by cotreatment with BMS-754807. Two independent experiments were performed. The average of one experiment with 5 to 7 replicates is shown. F, sensitization of NVP-TAE684 by cotreatment with BMS-754807. Two independent experiments were performed. The average of one experiment with 5 to 7 replicates is shown.

Figure 5

ACME analysis suggests vulnerabilities of a specific genetic context of KRAS-mutant CCLs to combination treatments. A, cartoon example of multiple hotspots for a single CCL cluster. B, the 10 compounds probed in the combination screen in LS513 cells and the rationale for their choice. Five were identified by ACME analysis. C, results from combination screen. Two biological and two technical replicates of the high-throughput screen were performed. The averages from the two biological replicates are shown. The synergy threshold is −0.17, and the antagonism threshold is +0.23. D, validation of the synergistic combinations identified. Two independent experiments were performed. The average of one experiment with 5 to 7 replicates is shown for each combination. E, AUC comparison of BMS-754807 and selumetinib in KRAS^G12-mutant CCLs in lung and large-intestine lineages.