4.4.6. Transwell Assay

The assay was performed according to previous protocols [17]. Cells were cultured in a Transwell® cell culture chamber (8 mm pore size; Corning, Lowell, MA, USA) at a density of 2 × <sup>10</sup>4/well. For a 24-h invasion assay, the chambers were first coated with MatrigelTM. A cell suspension in serum-free DMEM was cultured in the upper chamber of the Transwell® insert with or without different doses of genistein and solvent control. The lower part was filled with DMEM containing 20% FBS as a chemoattractant. After 24 h, cells that migrated or invaded into the lower surface of the membrane were fixed with 4% PFA and stained with crystal violet solution. Images of five random fields were captured by an inverted microscope. Cell migration/invasion was subsequently assessed using Image J software. Cell migration/invasion was calculated (as a percentage) from the relative numbers of cells in the genistein-treated and control groups.

#### 4.4.7. Western Blotting Assay

The assay was carried out according to previous protocols [17]. Briefly, cells were grown with or without different concentrations of genistein (12.5–100 μM) and solvent control for 30 min. Cells were then harvested and lysed in ice-cold cell lysis buffer. After protein extraction, the total protein concentration was quantified using standard protocols. Total protein (20 μg of protein/lane) was then separated on a 10% SDS-PAGE gel by electrophoresis. Separated proteins were transferred onto PVDF (Polyvinylidene Fluoride) membranes (BioRad, USA) and subsequently probed with the respective primary antibody followed by an HRP (Horse Radish peroxidase) -conjugated secondary antibody. Finally, the blots were developed by ECL (Electrochemiluminescence).

#### 4.4.8. RNA Sequencing

RNA sequencing was carried out as previously reported [62,63]. Cells were grown with DMSO and genistein (50 μM) for 5 h. Total RNA was isolated using the RNeasy Plus Mini Kit, following the manufacturer's instructions (Qiagen, Germantown, MD, USA). The quality of the RNA was first confirmed, and then the RNA was sequenced using the Illumina Hiseq X ten platform in GeneChem. Data from the sequencer were first subjected to quality control using FastQC and trimmed using trimgalore. Data processing included (i) trimming the Illumina adapter sequence and low-quality bases (phred score < 20) at the 3 end; (ii) discarding the reads with a length shorter than 20 (the paired reads were removed if any of the two reads did not meet the minimum length). Duplicates were then removed using Picard. For each sample, we counted the reads of individual transcripts using htseq-count. Differential analysis between treatments (genistein) and the control was performed using a count-based method, limma, implemented in R, and voom for normalization [64,65]. Significantly expressed genes were first screened for BH-adjusted *p* values less than 0.05 [66] and further filtered using a 2-fold-change minimum boundary (up- and downregulated genes labeled in the volcano plot). In parallel, we used GSEA v3.0 (Broad Institute, PreRanked mode) for enrichment analysis. To ensure consistency in our method for identifying significant genes, we used the t-statistic output from the limma as a metric for ranking. Here, 1000 gene set permutations were set as default, and gene sets

were obtained by collecting pathways from KEGG and biological processes from GO. A gene set with an FDR q value less than 0.05 was considered significantly enriched. For the heatmap, log2 transformed FPKM values (fragments per kilobase of transcript per million mapped reads) of the significant genes were used as input for heatmap generation.
