YWHAG Deficiency Disrupts the EMT‐Associated Network to Induce Oxidative Cell Death and Prevent Metastasis

Abstract Metastasis involves epithelial‐to‐mesenchymal transition (EMT), a process that is regulated by complex gene networks, where their deliberate disruption may yield a promising outcome. However, little is known about mechanisms that coordinate these metastasis‐associated networks. To address this gap, hub genes with broad engagement across various human cancers by analyzing the transcriptomes of different cancer cell types undergoing EMT are identified. The oncogenic signaling adaptor protein tyrosine 3‐monooxygenase/tryptophan 5‐monooxygenase activation protein gamma (YWHAG) is ranked top for its clinical relevance and impact. The cellular kinome and transcriptome data are surveyed to construct the regulome of YWHAG, revealing stress responses and metabolic processes during cancer EMT. It is demonstrated that a YWHAG‐dependent cytoprotective mechanism in the regulome is embedded in EMT‐associated networks to protect cancer cells from oxidative catastrophe through enhanced autophagy during EMT. YWHAG deficiency results in a rapid accumulation of reactive oxygen species (ROS), delayed EMT, and cell death. Tumor allografts show that metastasis potential and overall survival time are correlated with the YWHAG expression level of cancer cell lines. Metastasized tumors have higher expression of YWHAG and autophagy‐related genes than primary tumors. Silencing YWHAG diminishes primary tumor volumes, prevents metastasis, and prolongs the median survival period of the mice.


Supplemental Material and Methods
Immunoblotting.Cells were lysed using ice-cold RIPA buffer (Sigma Aldrich, USA), and protein was quantified using Braford Assay (Bio-Rad Laboratories, California, USA).Protein extracts were resolved by SDS-PAGE and electrotransferred onto nitrocellulose membranes (Merck & Co., New Jersey, USA).
The membranes were blocked in 1X Odyssey Blocking buffer (LI-COR Biosciences, Nebraska, USA) for 1 h at room temperature.The membranes were then incubated with the respective primary antibodies in 1X Odyssey Blocking buffer (Table S4) overnight at 4°C.Membranes were washed three times with TBST (1 X TBS, 50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.01% Tween-20) and further incubated with the respective IRDye® 680-conjugated or 800-conjugated anti-IgG secondary antibodies (LI-COR Biosciences USA) in 1 X Odyssey Blocking Buffer for 1 h at room temperature.Next, the membranes were washed with TBST with 0.01% SDS before analysis using an ODYSSEY CLx Infrared Imaging system (LI-COR Biosciences, USA).
Immunoprecipitation. Cells were lysed using ice-cold RIPA buffer (Sigma Aldrich, USA).The lysate was then incubated with anti-14-3-3γ antibodies (Abcam, UK) overnight at 4°C with constant rotation.Anti-14-3-3γ antibodies were then affinity precipitated using protein A/G beads (Santa Cruz, California, USA).The cell pellet was washed three times with PBS before resuspension in Lamelli's buffer (Bio-Rad Laboratories, USA).Proteins were released by boiling the samples for 10 min.Immunoblotting analyses were carried out as mentioned above, and the antibodies used are described in Table S4.
siRNA knockdown.SMARTpool ON-TARGETplus™ siRNA, which consists of a mixture of four siRNAs targeting YWHAG, YWHAE and YWHAH, was purchased from Dharmacon (Thermo Fisher Scientific, USA).Cancer cells were seeded in 96-well plates at a concentration of 300 cells/well prior to siRNA transfection with Lipofectamine 2000 (Invitrogen, Massachusetts, USA) according to the manufacturer's instructions.The siRNA-treated cells were incubated for 8 h at 37°C before the medium was replaced with complete medium for an hour and further incubated in the respective medium for 48 h.RNA extraction and real-time PCR.Total RNA was extracted using TRIzol® Reagent (Thermo Fisher Scientific, USA) followed by Pure NA.Fastspin Total RNA Extraction Kit (Research Instruments, USA) according to the manufacturer's protocol.Total RNA was quantified based on the A260/280 absorbance using a Nanodrop ND1000 (Thermo Fisher Scientific, USA).Total RNA was treated with DNaseI (Thermo Fisher Scientific, USA) and reverse transcribed using qScript cDNA SuperMix (Quantabio, Massachusetts, USA) according to the manufacturer's instructions.Quantitative PCR (qPCR) was performed in triplicate using the KAPA SYBR FAST qPCR master mix (Kapa Biosystems, Massachusetts, USA) and the C1000 Thermal Cycler with CFX96 Real-Time System module (Bio-Rad Laboratories, USA).Human and mouse primer sequences are listed in Table S5.Please refer to Supporting File 2 for the MIQE checklist for the reporting of our qPCR experiments.

FACS Apoptosis Assay.
Cells were detached from the plates before washing with ice-cold 1x PBS twice followed by resuspension in 1x binding buffer (BD Biosciences, USA).The cells were stained concurrently with 5 µL of fluorescein isothiocyanate (FITC) Annexin (BD Biosciences) and 5 µL of propidium iodide (PI) (BD Biosciences, USA) for 15 minutes at room temperature.The cells were washed again with ice-cold 1x PBS before resuspending in 400 µL of 1x binding buffer.Flow cytometry was performed using BD Accuri C6 Plus, and analysis was performed using FlowJo X software.
DESeq2 was used to analyse the differential gene expression between the control and treatment groups.Principal component analysis (PCA) was used to visualize the clustering pattern of the datasets.Metabolic genes were retrieved from Gene Ontology, EMT genes were obtained from EMTome, and autophagy genes were obtained from the Human Autophagy Database (HADb).VISEAGO was used to identify and group genes by their biological processes [2].The YWHAG-EMT interactome was constructed using the functional enrichment outputs according to a published method [3].The sequencing data of MKN74 cells with EMT induction have been deposited in the GEO database (GSE204929).Ingenuity Pathway Analysis (IPA) software (Qiagen, Germany) was employed to carry out the analysis of pathways and map downstream signalling molecules and possible interactions between 14-3-3γ and proteins involved both directly and indirectly in autophagy [4].
Cytoscape analysis.ClueGo is a plug-in on Cytoscape that combines gene ontology terms and Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathways to allow for an integrated network analysis [5].ClueGo was employed in this study to cluster the 14-3-3 interactomes isolated from BioPlex into their respective ontology terms.CytoHubba was utilized to compute the maximal clique centrality (MCC) scores and rank the nodes in the different networks obtained during meta-analysis [6].Genes with the top 5 highest MCC scores were identified as the hub genes in the EMT interactome.
Databases.Four datasets consisting of different EMT induction methods were retrieved from the Gene Expression Omnibus (GEO) repository.Out of the four datasets obtained, one used TGF-β1 on PANC-1 and A549 (GSE90566), another one induced hypoxia in PANC-1 with CoCl2 (GSE82104) and two overexpressed Zeb1 in H538 (GSE75487) and Zeb2 in DLD-1 (GSE148823).
The Human Autophagy Database (HADb), which is the first Human Autophagy-dedicated Database with information about the human genes involved in autophagy, was used to retrieve a complete and up-to-date list of human proteins involved both directly and indirectly in autophagy as described in the literature.
EMTome is a database that integrates profiles across multiple subtypes of cancers and was selected from The Cancer Genome Atlas (TCGA), which comprises the gene signatures, multiomics features, immune profiles and interactomes of EMT [7].EMTome was examined to determine the gene signatures involved in EMT and subsequently aided in the analysis for RNA-Seq and meta-analysis.
BioPlex is a database that was used to predict protein-protein interactions of 14-3-3 and the other EMT and autophagic proteins in cancer cells by utilizing datasets from collaboration between Harvard Medical School and Biogen [8].Predicted proteins obtained from the BioPlex interactome were used to unravel the interactome of 14-3-3 during EMT in cancer cells.
ANIA: Annotation and Integrated Analysis of the 14-3-3 interactome.ANIA was used to examine the possible human 14-3-3-binding phosphoproteins and phosphosites by utilizing datasets from studies that examined the phosphoproteins binding to 14-3-3 as well as high-throughput 14-3-3 affinity capture together with mass spectrometry-based studies in identifying potential targets [9].Proteins involved directly and indirectly in autophagy were used to examine their possible interactions with 14-3-3 by immunoprecipitation.
Survival Outcome Analytics using Clinical Databases.To assess the clinical relevance of YWHAG levels on patient survival, we interrogated four cancer cohort databases.First, by using The Cancer Genome Atlas Pan-Cancer Clinical Data Resource (TCGA-CDR) and Gene Expression database of Normal and Tumor tissues 2 (GENT2), YWHAG expression was examined in deposited tumor tissues as well as in cognate normal tissues.Next, we stratified the cancer patients into high and low YWHAGexpressing tumors.In the PrognoScan database [10], cohort studies that reported overall, distant metastasis-free, disease-specific and relapse-free survival and had significantly corrected P values were examined.The Prediction of Clinical Outcomes from Genomic Profiles (PRECOG) [11] evaluates the expression of 23,287 genes across 39 cancer types and clinical outcome data from 166 published expression datasets covering 18,000 tumors, thus extensively studying the associations between genomic profiles and cancer outcomes.

Human tumor biopsies.
Human frozen tissue biopsies from the head and neck, breast, stomach and colon were purchased from Proteogenex (Proteogenex, California, USA) (Table S6).HepG2 siYWHAG ) at 48 h after siRNA transfection.TBP and 18S were used as housekeeping genes.
(B) Representative immunoblots of 14-3-3 family members in wild-type and YWHAG-knockdown cancer cells.The β-tubulin protein was used as a loading and transfer control from the same samples.

Figure S4. YWHAG knockdown ceases active EMT in MKN74.
(A) Schematic diagram of the experimental setup to study the effect of YWHAG knockdown in MKN74 undergoing active EMT.EMT was induced using DMOG (0.5 mM) or TGF-β1 (10 ng/mL) for 24 h.
Successful EMT was confirmed by the upregulation of mesenchymal-associated genes and downregulation of epithelial genes.After 24 h of EMT induction, the cells were transfected with 10 nM siYWHAG and maintained for another 24 h to examine the effect of YWHAG knockdown on mesenchymal and epithelial gene expression.Three control groups were carried out concurrently, including untreated cells, EMT cells, and MKN74 siYWHAG at 0 h.

(B- C )
Representative brightfield images (C) and Euclidean distance (D) of untreated, DMOG-and TGF-β1-treated MKN74 and MKN74 siYWHAG transfected at 0 h or 24 h.Scale bar = 100 M.(D) Gene expression of epithelial and mesenchymal markers in DMOG-or TGF-β-induced EMT in MKN74 (left), MKN74 siYWHAG transfected at 0 h (mid) and MKN74 siYWHAG transfected at 24 h (right).The blue dotted line represents the relative expression of untreated cells.TBP and 18S were used as housekeeping genes for qPCR.Data, wherever applicable, are represented as the mean ± SD from at least three independent experiments.***p < 0.001, **p < 0.01, *p < 0.05 and n.s.denotes not significant (Mann-Whitney U test).

Figure S6 .
Figure S6.High YWHAG is a common feature in many tumor types.(A-D) Tumor type (red) which showed statistically significant higher YWHAG mRNA level (A, C) or which did not show significant differences (B, D) compared with cognate normal tissues (in gray) obtained from TCGA (A, B) and GENT2 (C, D) databases.TPM denotes transcripts per million.Values below the box-whisker plots represent the number of patients from whom the tumour samples were derived.* p<0.01.n.s.denotes no significance.

Figure S7 .
Figure S7.Interpretation of the kinase inhibitor screen during EMT of MKN74 cells (A) Schematic diagram of the kinase inhibitor screens in MKN74 and MKN74 siYWHAG cells during EMT induction by DMOG or TGF-β.Illustration included the possible correlation between YWHAG dependency and the outcomes on migratory distance (EMT progression, EMT inhibition or no change).(B) Illustration of Euclidean distances between multivariate centroids of the cell populations at 0 h and 48 h in MKN74 and MKN74 siYWHAG cells treated with EMT inducters and kinase inhibitors.(C) Schematic diagram of the calculation and classification of kinase inhibitors in the screens.For every kinase inhibitor, the average Euclidean distance, which was measured every 4 h for 48 h after EMT induction, was considered the migratory distance.The log fold change in migratory distance between treated (with kinase inhibitor) and DMSO-treated MKN74 cells was calculated, and the temporal changes were plotted.The AUC, which reflects the overall effect of kinase inhibitors on EMT, was determined and compared between wild-type MKN74 and MKN74 siYWHAG cells.Kinase inhibitors were classified into EMT activators on YWHAG-dependent kinases or EMT inhibitors on YWHAGdependent/independent kinases based on the similarity/difference in AUC between genotypes.Examples of temporal changes in the log fold change of migratory distance for each classification are shown.

Figure
Figure S8.Protein-protein interactomes of 14-3-3 family members.(A)Table shows the number of interacting proteins with the different isoforms of 14-3-3 isoforms derived from the BioPlex Interactome database.(B)Venn diagrams showing common interacting proteins among the different 14-3-3 isoforms.The number of interactomes for each 14-3-3 isoform is indicated in parentheses.(C)Gene ontologies of the interactomes of 14-3-3.The size and color in circle correspond to relative number of interacting proteins in the gene ontologies of the same color.

Figure S9 .
Figure S9.YWHAG deficiency induces oxidative stress in MCF7 cells.(A) Intracellular ROS profiles of MKN74 and MCF7 transfected with different concentrations of siYWHAG (0, 10 nM and 100 nM) with or without the co-treatment of EMT inducers, DMOG-and TGFβ.(B) Graph showing the kinetics of intracellular ROS (left) and cell viability (right) in untreated, DMOGand TGF-β-treated MCF7, MCF7 siYWHAG , MCF7 siYWHAG (10nM) and MCF7 siYWHAG (100nM) cells over a period of 48 h at 8-h intervals.(C-D) Relative mRNA and protein expression of oxidative stress markers in MCF7 cells treated with DMOG and TGF-β.TBP and 18S were used as housekeeping genes for qPCR.The blue dotted line represents the relative expression of untreated cells.(E) Relative mRNA expression of oxidative stress markers in MCF7 siYWHAG (10nM) and MCF7 siYWHAG (100nM) cells.In (C) and (E), the blue dotted lines represent the relative expression of untreated cells.Data are represented as the mean±SD from at least three independent experiments.***p < 0.001, **p < 0.01, *p < 0.05 and n.s.denotes not significant (Mann-Whitney U test).

Figure S10 .
Figure S10.YWHAG deficiency in HaCaT cells.(A) Relative fold change in YWHAG mRNA levels in HaCaT and HaCaT siYWHAG cells.(B) Immunoblot analysis of 14-3-3γ and the other 14-3-3 isoforms in HaCaT and HaCaT siYWHAG cells.(C) Apoptosis assay by flow cytometry using Annexin V-FITC and PI double staining in HaCaT and HaCaT siYWHAG cells.(D) Cell viability of HaCaT and HaCaT siYWHAG cells over a period of 48 h at 8-hour intervals.Data, wherever applicable, are represented as the mean  SD from at least three independent experiments.***p < 0.001, **p < 0.01, *p < 0.05 and n.s.denotes not significant (Mann-Whitney U test).

Figure S13 .
Figure S13.Expression of YWHAG and selected autophagic markers in human tumours.Relative fold change in mRNA (top) and protein (bottom) of autophagy markers and YWHAG in breast, colorectal, gastric, and head and neck tissues at various tumour stages (stages I, II and III; metastatic M).TBP and 18S were used as housekeeping genes for RT-PCR.Representative immunoblots are shown.β-tubulin was used as a loading and transfer control for immunoblot analysis from the same samples.

Table S1 . Patients with high YWHAG-expressing cancers (TCGA database) have a shorter median survival time #
Median survival cannot be determined because it did not reach 50% survival probability.The number of patients is indicated in parentheses.Tumor types with <200 cancer patients were underpowered and not included in the analysis.

Table S2 . Ten cohorts from the PrognoScan database showed that high YWHAG expression was associated with poor prognosis.
Ten cohort studies covering various types of cancers and end points from the PrognoScan database illustrated the hazard ratio of cancer-specific events.Observations demonstrated that high YWHAG expression is correlated with poorer prognosis in patients.HR, hazard ratio; CI, confidence interval.

Table S3 . Global unweighted meta-Z score of all cancers from the PRECOG database
The PRECOG database encompasses 166 cancer expression datasets, including overall survival data for ~18,000 patients diagnosed with 39 distinct malignancies.