Fig 2: EZH2 inhibition derepresses NANOG target gene through downregulation of repressive histone marks (H3K27me) and increases activation histone marker (H3K27ac)(A) FACS measurement of ROS production in Huh7 cells. Both FAOi and EZH2i promoted mitochondrial ROS levels and were compared for single, or combination FAOi and EZH2i treatments.(B) Loss of ARID1A enriched label retaining (slow-cycling) cell subpopulation and reduced mitochondrial ROS production. Huh7 cell stably expressing control shRNA or shRNA against ARID1A were treated with vehicle, Etomoxir, GSK126, or combination of Etomoxir and GSK126. Cells were stained with CellTrace Violet (label retaining population for which intensity is reduced at every cell division) and DiOC6 (mitochondria membrane potential) and subjected to flow cytometry analysis. (Right) Quantification of percentage of Dioc6 positive population (Top panel) and quantification of percentage of CellTrace Violet positive but Dioc6 negative population (Bottom panel) in cells treated with sh-scrambled versus ARID1A-scilenced Huh7 cells. Purple-colored (Top) or Red-colored quadrants (Bottom) were compared in FACS graphs.(C) EZH2 inhibition derepresses NANOG target through downregulation of repressive histone marks (H3K27me) and increases activation histone marker (H3K27ac).(D) COX6A2 gene was upregulated upon FAOi and EZH2i treatment. Huh7 cells were treated singly or in combination with FAOi and EZH2i. RT-qPCR was performed to detect mRNA level of COX6A2 after FAOi and EZH2i treatments. *: Statistical significance (p < 0.05 by Student’s T-test).(E) EZH2 inhibition derepress OXPHOS NANOG target and increases activation histone marker (H3K27ac). ChIP-qPCR was performed to show the fold enrichment of anti-H3K27Me3 (Left) or anti-H3K27Ac binding (Right) on COX6A2 promoter region. The site examined corresponds to the promoter proximal NANOG binding site of the COX6A2 gene. *: Statistical significance (p < 0.05).(F) ChIP-qPCR was performed to show the fold enrichment of anti-NANOG, anti-EZH2, or anti-EED binding on COX6A2 promoter region. Stars denote statistical significance (p < 0.05, Student T-test).(G) Hypothetical model of NANOG-mediated gene suppression through NANOG-PRC2 interactions. PRC2 complexes and PPARd activators compete NANOG tryptophan (W)-rich domain of NANOG for global transcriptional suppression or activation. PRC2 component EED and ubiquitin E3 ligase FBXW8 compete NANOG phosphodegron sequence PEST domain (P, E, S, and T-rich phosphodegron sequence) for stabilization or E3 ligase FBXW8-dependent degradation. Inhibitors of NANOG-PRC2 interactions with compete NANOG phosphodegron domain (PEST domain) to destabilize NANOG protein to inhibit self-renewal abilities of TICs and reduce tumor growth in humanized mice. Therefore, PRC2 components stabilized and differentially regulated NANOG target genes which depend on the NANOG W domain-binding partners (PPARd or EZH2/SUZ12). These mechanisms would be conserved in many different TICs or embryonic stem cells.

Fig 3: PRC2 component proteins (EED, SUZ12, and EZH2) interact with NANOG to stabilize NANOG(A) Schematic of NANOG deletion constructs containing an N-terminal Flag tag. These deletion mutants of NANOG were tested for EED binding.(B) HEK 293T cells were cotransfected with the indicated Myc-tagged EED full-length constructs and various Flag-tagged NANOG deletion constructs as listed in (A) and as blot lane numbers. Cell lysates were immunoprecipitated with anti-Myc or anti-Flag followed by immunoblotting with anti-Flag, as indicated. Nanog interacts with PRC2 components EED through carboxyl terminal domains (C1-W domains). Note NANOG with C1-W domain loses this interaction.(C) Reciprocal IP-western blot analyses confirmed interactions between NANOG and EED. HEK 293T cells were cotransfected with the indicated Myc-tagged NANOG deletion constructs and Flag-tagged EED full-length constructs. Immunoprecipitation and immunoblotting were conducted with antibodies as indicated in the immunoblots. Different deletion mutants of NANOG were tested for EED binding. Note that NANOG N-terminus and C1-W domain is needed for interaction with EED.(D) The EED silencing reduced NANOG protein levels in TICs. Student’s T-test was used for statistical analyses. Star marks denote significantly different (p < 0.05).(E) RT-qPCR analyses of TICs with EED-knockdown. Silencing of EED did not alter NANOG mRNA levels (Left) while silencing of EED induced COX6A2 expression (Left), but reduced OCT4 and SOX2 mRNAs.(F) Spheroid colony formation assay. Upper-EED expression transformed p53-deficient hepatoblasts and was potentiated by NANOG. Lower-HNF4A transcription was unaffected by NANOG; however, EED repressed HNF4A transcription. Stars denote statistical significance (p < 0.05).(G) EED silencing reduced protein stability of NANOG. Huh7 cells were transduced with scrambled shRNA (sh-scr) or shRNA targeting EED. Seventy-two hours after transduction, cycloheximide (CHX) was added and cells were harvested at the indicated times. NANOG protein level was detected by immunoblotting (Top panel). Results of a representative experiment (n = 3) are plotted as percentage of starting NANOG protein level for half-life determination (Bottom panel).(H) IP-western blots of EED binding to NANOG mutants. CD133(-) non-TICs and TICs were lysed and immunoprecipitated with anti-NANOG antibody and examined for immunoblot analyses. Endogenous EED interacts with NANOG. Huh7 cells were transfected with shRNA for EED followed by immunoblotting with anti-NANOG or anti-EED as indicated. Cells silenced for EED did not co-precipitate NANOG as was observed for control sh-scrambled treated cells. As shown, TICs associated with the three subunits of PRC2 (EED, EZH2, and SUZ12), RbAp46/48, JARID2, and AEBP2.(I) Hypothetical mechanisms of EED-NANOG interaction promoting stabilization and activation of NANOG signaling in TICs. EED may stabilize NANOG through its dependence on PEST domain phosphorylation allowing homodimerization. Different deletion mutants of NANOG were tested for EED binding. Note that NANOG N-terminus and C1-W domain is needed for interaction with EED. Created with biorender.com (Agreement Number: CN259LCOTO).(J) (Left) sh-EED and sh-Scrambled Huh7 cells were transfected with flag-tagged NANOG vector for 48 h. At 24 h post-transfection, some cells were treated with 10 µM MG132. Whole cell lysates were used in western blot to quantify Flag-tagged NANOG. EED silencing reduced NANOG protein levels. Proteasome inhibitor (MG132) treatment enhanced NANOG protein levels that were accentuated by EED silencing. (Right) Immunoprecipitation-western blot analysis of NANOG in EED knockdown Huh7 cells. The sh-EED and sh-scrambled-transduced Huh7 cells were lysed with RIPA buffer. The cell lysates were pre-cleared with magnetic Protein A beads then incubated with anti-Nanog antibody overnight. The immunocomplex was pulled down with Protein A beads followed by western blot analysis of NANOG. Blot was prepared and analyzed by ImageJ.(K, Top) The immunocomplex was pulled down with Protein A beads followed by western blot analysis of FBXW8. Immunoprecipitation-western blot analysis of FBXW8 in EED knockdown Huh7 cells. Co-IP western blot analysis of NANOG and its PEST domain-associated proteins. Sh-EED and sh-scrambled Huh7 cells transfected with Flag-NANOG vector were treated or untreated with 10 µM MG132. The cell lysates were pre-cleared with magnetic Protein A beads then incubated with anti-Flag antibody overnight. The immunocomplex was pulled down with Protein G beads followed by western blot analysis of FBXW8 and PIN1 proteins. Western blot analysis of FBXW8, ubiquitin, and PIN1 was done to determine their association with NANOG under these conditions. Sh-EED and sh-scrambled Huh7 cells co-transfected with Flag-NANOG and MYC-FBXW8 vector were treated or untreated with 10 µM MG132 (proteasome inhibitor). Following the same immunoprecipitation procedure, the lysate was immunoblotted with anti-Myc antibody. IP Flag-NANOG/WB FBXW8 is for detection of FBXW8. The next line is for WB PIN1. (K, Bottom) The same experimental procedure as described above. EZH2, EED, and ß-Actin proteins were detected by western blot analysis.(L) EED knockdown and sh-scrambled Huh7 cells were cultured, and some were treated with 10 µM MG132 for 24 h. Cells were lysed and total RNA collected. The RNA was reverse transcribed to cDNA and subsequently used in quantitative-PCR with SYBR green. Oligonucleotides for NANOG and COX6A2 were used to amplify the respective gene.(M) EED-NANOG cooperation in tumor development. Tumor growth after TIC subcutaneous transplantation in NSG mice was markedly suppressed by EED KD but partially rescued by concomitant expression of NANOG. NANOG alone enhanced tumor growth which was attenuated by EED KD below the growth observed in control TICs.(N) EED KD reduced NICD-induced tumor growth by p53-/- hepatoblasts orthotopically transplanted into the livers of NSG mice.(O) Screening for selective inhibitors for SUZ12-NANOG interaction. Diagram of drug screening by fluorescence polarization assays.(P) XTT cell viability assays in CD133(+) and CD133 (-) Huh7 cells. Note NSC8090, NSC14540, and NSC123127 (Doxorubicin) are selectively cytotoxic to CD133+ cells. Small molecule NSC8090 selectively kills CD133+ Huh7 cells. “: p < 0.05.(Q) NSC8090 blocks EED-NANOG interactions in CD133+ Huh7 cells.(R) IC50 values of combination treatments of a PRC2-NANOG inhibitor and/or sorafenib were examined in different HCC cell lines in the presence or absence of ARID1A mutations. Stars denote statistical significance.

Fig 4: PRC2 complexes and PAPRd competes the W-rich domains of NANOG(A) (Upper Left) Schematic of PPARd truncated expression constructs with c-Myc epitope tag. NANOG binding activity is indicated. (Lower Left) Hypothetical model of FAO inhibitor-mediated apoptosis in TICs. NANOG promotes FAO in HCC cells. (Upper Right) IP-western blot analyses between NANOG and PPARd truncation mutants. The N-terminus of NANOG (aa1-72) binds PPARd. (Lower Right) Confirmation of PPARd truncation mutant expression. Lane numbers correspond to constructs shown in schematic. Numbers on the top of immunoblots denote which PPARd mutant was combined with full-length NANOG (FL).(B) Left-diagram of Flag-NANOG constructs used for testing PPARd binding activity. Right- Co-IP combinations with NANOG mutants and Flag-PPARd mutants; lane numbers correspond to latter mutants. IP-western blots showed that NANOG and PPARd interacted through W domains of NANOG. Numbers on the top of immunoblots denote which NANOG mutant was combined with full-length PPARd (FL).(C) Hypothetical mechanisms of EED-NANOG interaction promoting stabilization and transcriptional suppression by NANOG-PRC2 interactions in TICs. EED may stabilize NANOG through its dependence on PEST domain phosphorylation. PKCe-mediated phosphorylation of T200 of NANOG promotes homodimerization. Transcriptional suppressor PRC2 complexes (EZH2-SUZ12) and transcriptional activator PPARd-NANOG share binding-sites and possibly compete with W domain of NANOG.(D) Hypothetical model of FAO inhibitor-mediated apoptosis in TICs. NANOG suppresses mitochondrial respiration by suppressing OXPHOS genes, but transactivates fatty acid oxidation genes (FAO) by transcriptional activation through competing out PRC2 complex binding in tryptophan (W)-rich domain that is used for NANOG dimerization and activation as well. The TCA cycle generates NADH. The transfer of reducing equivalents from NADH to NADP+ via nicotinamide nucleoside transhydrogenase (NNT) provides the reducing potential energy (NADPH). NADPH reduces glutathione and controls mitochondrial ROS. (Right) Inhibition of FAO inhibits energy production in TCA cycle and OXPHOS and promotes ROS production, leading to cytochrome c release and apoptosis. Loss of mitochondrial membrane potential ?? promotes apoptosis of TICs via BAX/BAK oligomerization.(E) Changes in HNF4A and c-Myc mRNAs in response to ARID1A-1 or ARID1A-2 silencing. RT-qPCR analysis for ARID1A knockdown in Huh7 cells decreased ARID1A and HNF4A but increased Myc mRNA level. Stars denote statistical significance (p < 0.05, Student T-test).(F) Cytosolic and mitochondrial marker expression. Left-Huh7 cells treated with shRNA against scrambled or ARID1A were treated with FAOi and EZH2i for 1 h, then cell lysates were harvested at time points as indicated. Cells were fractionated into cytoplasmic (left panel) and mitochondrial fractions (right panel), followed by SDS-PAGE and immunoblotting with antibodies against cytC, Bak, Bax, Cu/ZnSOD, or VDAC, as indicated.(G) Knockdown of ARID1A alleviated apoptotic activity of FAOi and EZH2i. Protein expression levels for cytochrome c and BAX were measured. The densitogram of immunoblots shown in panel E (left) was quantified by ImageJ.

Fig 5: CRISPR-Cas9 screenings identified synthetic lethality co-mutations in ARID1A and PRC2 components(A) GeCKO lentivirus-library screening identified functional loss of genes that promote alcohol-mediated HCC development. (Middle) Lentiviral vector diagram for Cas9 and sgRNA for genome-scale knockout of coding sequences in human cells.(B) Identification of candidate genes needed for metastases.(C) Mutational frequencies of chromatin remodeling genes in alcohol- or/and HCV-associated HCCs.(D) GSEA identified PRC2 components (EZH2, EED, and SUZ12) were coenriched on NANOG-bound gene regions. (Bottom) Venn diagram of GEO datasets of HepG2 cells with wild-type ChIP-Seq data from mice liver TICs showed OXPHOS genes were coenriched with ARID1A and Nanog in ARID1A and SNF overexpressing datasets examining NANOG-bound regions. Data comparison from integrated analyses of ARID1A and SNF ChIP-Seq from HepG2 and NANOG ChIP-Seq data of TICs isolated from mice with alcohol-associated HCCs with ARID1A mutations.(E) Overlap of RNA profiling and Nanog ChIP-Seq data for NS5A transgenic and chronic alcohol-treated mouse liver TICs with ARID1A mutation showed PRC2 components (EZH2, EED, and SUZ12) were coenriched in NANOG-bound regions. Gene Set Enrichment Analyses (GSEA) identified PRC2 components EED, EZH2, and SUZ12 were co-enriched in NANOG-bound regions.(F) Genome-scale knockout, negative selection screening (GeCKO) in ARID1A-mutant HCCs that are resistant to sorafenib treatment. (Right) GeCKO screening for metastasis with single-guide RNAs (sgRNAs).(G) Synthetic lethality candidate genes (PRC2 genes) are listed.(H) Silencing ARID1A increased NANOG mRNA levels in TICs, but not in primary hepatocytes. Lipopolysaccharide (LPS) stimulation induces NANOG in TICs and sh-ARID1A-transduced primary hepatocytes, but not in primary hepatocytes. One of the major HCC risk factors is alcoholism that allows endotoxin leakage from intestinal tracts, leading to high endotoxin levels, stimulating NANOG transactivation. As LPS tolerance is evaded in TICs (not in primary hepatocytes), LPS stimulation induces NANOG. Star denotes statistical significance (p < 0.05, by Student T-test).(I) Hypothetical model of generation of TICs. Cancer-promoting mutations (ARID1A) increase PRC2 complex activity (including EZH2 component) and induce HCV/alcohol-mediated stem cell program (slow growing cells), leading to TIC-initiated HCC development.(J) Expression of HNF4A mRNA is correlated with ARID1A mRNA, but inversely correlated with PRC2 components (EZH2, EED, and AEBP2) in patients with HCC. RNA-seq from TCGA HCC patients were analyzed for the correlation of the expression between of HNF4A and ARID1A and that between HNF4A and PRC2 components EZH2, EED, and AEBP2.(K) Hypothetical model for interactions between ARID1A and PRC2 regulating TIC development. Alcohol intake changes gut microbiota to promote Gram(-) bacteria overgrowth in intestinal tracts and leaky gut, leading to endotoxin leakage from intestinal tracts. Endotoxin (LPS) in blood stream stimulates TLR4-CD14 complexes to transactivate NANOG to promote stem cell program. PRC2 complexes activate stemness gene and suppress anti-stemness genes (OXPHOS genes).