Fig. 3: Pooled screens reveal metastasis-specific vulnerabilities.

a The analysis workflow of pooled shRNA screens⁶¹ in primary and metastatic cells. The screens are quantified by microarrays, after which all measurements are extracted, and each measurement is annotated with gene, shRNA, cell line, time point, replicate and primary/metastasis information. The siMEM hierarchical regression algorithm⁶² is then applied to test for differences between primary and metastasis cells. b Volcano plot illustrating significant predictions from the primary vs. metastasis analysis, including several components of the Notch pathway (red). c The abundance of NOTCH3 shRNA is reflected by the fluorescence intensity in the Affymetrix array measured across the 3 time points of the screen (0, 1 and 2, corresponding to T0, T1 and T2, respectively). For each patient there are 2 plots, first showing the result from the primary tumor sample and the second showing the result from the metastatic sample from the same patient. Each time point in each of the plots is represented by 3 replicates. d Notch3 essentiality scores across a set of 29 cell lines (see Supplementary data 5 for details) are shown. The x axis (labeled index) represents the ordered list of the cell lines shown in the plot. Samples are paired based on the patient they were obtained from, and each pair is connected by a gray line. Cell lines derived from the primary tumors are contoured with a blue line; cell lines derived from the metastatic samples are contoured with red lines. “Pri” primary tumor derived cell lines, “met” metastases derived cell lines, “HPV+” HPV positive cell lines, “rec” recurrence derived cell lines. Insert: Distributions of Notch3 essentiality scores of primary tumors derived cell lines (blue) and metastases derived cell lines (red) are shown on a density plot that shows frequency of each score in each cell line. For cell lines derived from metastatic samples, the distribution of essentiality scores is bimodal, allowing to separate the cell lines into 2 groups (see red contours). Comparison between these 2 groups showed a significant difference (Welch Two Sample t-test, p = 4.029e−05); additionally, the bottom group of metastatic samples was significantly different from the primary tumors (Welch Two Sample t-test, p = 1.555e−06). e Metastatic and primary lines from each set were labeled by infecting with lentivirus expressing either H2B-GFP or H2B-RFP, mixed in an equal ratio (300 cells/well of each line), plated in 384 well plates, and transfected with Dharmacon siRNA pools (using RNAmax Lipofectamine) against the target of interest. Fluorescent images were acquired using In Cell analyzer and shown in Fig. S1. Competitive survival/proliferation is assessed by comparing the ratio of primary/metastatic cells as a function of time. The data is presented as mean ± SEM. f HNSCC lines were engineered to express (doxycycline) inducible shRNA against Notch3. Cellular growth upon induction of the shRNA against Notch3 was measured in primary and metastatic lines using Incucyte Zoom Instrument and the increase in confluency between day 5 and day 1 was calculated and normalized to control (untreated) cells for each cell line. The graph illustrates that knockdown of Notch3 selectively inhibits growth of metastatic cells. The data is presented as mean ± SEM.