TGFβ Promotes Genomic Instability after Loss of RUNX3

Cancer Research Molecular Cell Biology TGFb Promotes Genomic Instability after Loss of RUNX3 Vaidehi Krishnan1, Yu Lin Chong1, Tuan Zea Tan1, Madhura Kulkarni1,2, Muhammad Bakhait Bin Rahmat1, Lavina Sierra Tay1, Haresh Sankar1, Doorgesh S. Jokhun3,4, Amudha Ganesan1, Linda Shyue Huey Chuang1, Dominic C. Voon1, GV Shivashankar3,4, Jean-Paul Thiery5,6, and Yoshiaki Ito1 Abstract matory cytokine expression and acquisition of the senescenceassociated secretory phenotype (SASP). Recapitulating the above findings, tumors harboring a TGFb gene expression signature and RUNX3 loss exhibited higher levels of genomic instability. In summary, RUNX3 creates an effective barrier against further TGFb-dependent tumor progression by preventing genomic instability. These data suggest a novel cooperation between cancer cell– extrinsic TGFb signaling and cancer cell–intrinsic RUNX3 inactivation as aggravating factors for genomic instability. Significance: RUNX3 inactivation in cancer removes an antioxidant barrier against DNA double strand breaks induced by TGFb expressed in the tumor microenvironment. Cancer Res; Introduction on cancers by activating antiproliferative signaling or by imparting pro-oncogenic properties (3). The contradictory facets of TGFb signaling and the molecular switches that enable this transition have been a subject of intense study. For example, in pancreatic cancers, TGFb induces Sox4 that converts TGFb from a tumorpromoting into a tumor-inhibiting factor by inducing apoptosis (4). In contrast, in lung cancers, TGFb functions as a tumorpromoting factor that induces angiogenesis, metastasis, and poorer patient survival, through the induction of EMT (5, 6). EMT is a central developmental process whereby cells lose their epithelial identity and gain mesenchymal features (7). TGFb regulates EMT by SMAD-dependent and SMAD-independent pathways such as the PI3K–AKT, ERK MAPK, p38 MAPK, and JUN N-terminal kinase (JNK). These intricate signaling pathways actively crosstalk and impart potent oncogenic phenotypes such as increased migration, invasion, drug resistance, and stemness (8). Despite extensive studies on the pro-oncogenic effects of TGFb, it remains unknown whether TGFb disrupts cancer cell genomic integrity. In an earlier work, TGFb was shown to control the DNA damage response (DDR) pathway by regulating ATM (9). Along similar lines, TGFb inhibition type I receptor led to reduced Chk2, Rad17, and p53 phosphorylation and Smad2 and Smad7 localized with DSB repair proteins (10). Recently, TGFb was shown to trigger aneuploidy and genomic instability in cells undergoing EMT by inducing mitotic abnormalities (11). Here, we identified that stromal TGFb may be involved in triggering another distinct form of genomic instability, specifically by the generation of oxidative DNA damage in cells deficient for the tumor suppressor gene, RUNX3. Genomic instability promotes the acquisition of the cancer phenotype by allowing mutational accumulation (1). In most hereditary cancers, genomic instability arises by the inactivation of DNA damage repair genes like BRCA1, BRCA2, TP53, or mismatch repair genes. In sporadic cancers, genomic instability arises due to cancer cell–specific defects like DNA replication stress, aberrant AID/APOBEC activity, or micronuclei-mediated chromothripsis (2). However, it has remained unexplored whether genomic instability can be fuelled in a cell-extrinsic manner by cytokines from the tumor microenvironment. In this regard, the TGFb is the most abundantly secreted cytokine by both tumors and their surrounding stromal cells (3). The secreted TGFb, in turn, might elicit paradoxical effects 1 Cancer Science Institute of Singapore, National University of Singapore, Singapore. 2Lee Kong Chian School of Medicine, NTU, Singapore. 3Mechanobiology Institute, National University of Singapore, Singapore. 4Department of Biological Sciences, National University of Singapore, Singapore. 5Yong Loo Lin School of Medicine, National University of Singapore, Singapore. 6UMR 1186 INSERM Institute, Gustave Roussy, Villejuif, France. Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). Corresponding Authors: Yoshiaki Ito, Cancer Science Institute of Singapore, National University of Singapore, MD-6, 14-Medical Drive, Singapore 117599, Singapore. Phone: 65-6516-2242; Fax: 65-6873-9664; E-mail: [email protected]; Jean-Paul Thiery, [email protected]; and Vaidehi Krishnan, [email protected] doi: 10.1158/0008-5472.CAN-17-1178 2017 American Association for Cancer Research. 88 Cancer Res; 78(1) January 1, 2018 78(1); 88–102. 2017 AACR. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Studies of genomic instability have historically focused on intrinsic mechanisms rather than extrinsic mechanisms based in the tumor microenvironment (TME). TGFb is the most abundantly secreted cytokine in the TME, where it imparts various aggressive characteristics including invasive migration, drug resistance, and epithelial-to-mesenchymal transition (EMT). Here we show that TGFb also promotes genomic instability in the form of DNA double strand breaks (DSB) in cancer cells that lack the tumor suppressor gene RUNX3. Loss of RUNX3 resulted in transcriptional downregulation of the redox regulator heme oxygenase-1 (HO-1 or HMOX1). Consequently, elevated oxidative DNA damage disrupted genomic integrity and triggered cellular senescence, which was accompanied by tumor-promoting inflam- DNA Damage during TGFb Signaling Materials and Methods Cell lines and cell culture A549, NCI-H1299, NCI-H292, PC-14, and NCI-H23 cells were obtained from ATCC. AZ-521 was from the Korean Research Institute of Bioscience and Biotechnology (KRIBB, Korea). HGC-27 was from Riken Cell Bank. Early passage cells were utilized for all experiments. Cells were cultured in RPMI medium (RPMI1640, Nacalai Tesque) supplemented with 10% (v/v) FBS (Biowest). RNA-Sequencing and raw data processing Samples were harvested and total RNA was extracted according to the manufacturer's instructions (Qiagen RNeasy Mini Kit, Qiagen). Samples were processed further for RNA-Seq analysis, as described under Supplementary Methods. Gene knockdown and plasmid transfection A549 cells were transfected with pooled siRNA (Dharmacon) as described in "Experimental Procedures" (Supplementary Information). Jet Prime transfection reagent (Polypus) was used for siRNA transfection into NCI-H1299, PC-14, NCI-H23, AZ-521, and HGC-27, according to the manufacturer's instructions. Immunofluorescence Cells were fixed with 4% paraformaldehyde for 15 minutes at room temperature, permeabilized with 0.5% Triton X-100 in PBS for 15 minutes and blocked using 2% BSA, 5% FBS in 0.1% Triton X-100 for 30 minutes. Antibodies diluted in 2% BSA in 0.1% Triton X-100 were incubated overnight at 4
C. Alexa Fluor– conjugated secondary antibodies (1:1,000) were added for 1 hour, at room temperature. Coverslips were mounted with Prolong Gold Anti-fade (Invitrogen) with DAPI. Images were captured using the Olympus FluoView FV1000 Confocal microscope using the 60 oil objective. Cells containing greater than 5 gH2AX foci or 53BP1 foci were considered as positive for DNA damage accumulation. gH2AX antibody was from EMD Millipore (catalog no. 05-636) and the 53BP1 antibody was from Abcam (catalog no. ab-36823). www.aacrjournals.org Senescence-associated b-galactosidase assay Senescence-associated b-galactosidase (SA-b-gal) assay was done according to the manufacturer's instructions (Abcam). Briefly, cells were washed twice with PBS, fixed with fixation buffer for 15 minutes at room temperature, and stained with b-galactosidase detection solution overnight at 37
C. Cells were imaged and blue cells were scored using the ImageJ (NIH, Bethesda, MD) software. Reactive oxygen species measurement by flow cytometry For detection of total reactive oxygen species (ROS) levels, cells were washed twice with PBS and incubated with 10 mmol/L carboxy-H2DCFDA (6-carboxy-20 ,70 -dichlorodihydrofluorescein diacetate) for 20 minutes or 5 mmol/L CellROX Deep Red Reagent (Molecular Probes) for 30 minutes. The FACS LSRII flow cytometer (Becton Dickinson) was used for data acquisition. Data analysis was performed using the FlowJO single-cell analysis software. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 The RUNX family of proteins is comprised of three heterodimeric transcription factors (TF), RUNX1, RUNX2, and RUNX3. Of these, RUNX1 and RUNX3 are inactivated in cancers by mutation or epigenetic deregulation, emphasizing their status as bonafide tumor suppressors (12). In the developmental context, RUNX proteins have emerged as critical modulators of TGFb signaling through their physical interaction with R-SMADs and with the bone morphogenetic protein (BMP)-specific SMADs and SMAD4. RUNX cooperates with SMADs to induce synergistic gene transcription of the cell-cycle inhibitor, p21, and the apoptosis inducer, BIM in a TGFb-dependent manner (13, 14). In pancreatic cancer, the SMAD4 gene dosage was recently reported to control RUNX3 transcription; in turn, the expression levels of RUNX3 control the switch between local proliferation and metastasis (15). RUNX3, in conjunction with TGFb, therefore plays key roles in dictating cancer progression. Here, we uncovered that TGFb promotes genomic instability in RUNX3-deficient cells by inducing oxidative stress–associated DNA damage by the downregulation of the redox regulator, HMOX1. By providing defense against TGFb-dependent promotion of cancer progression, our results unveil a unique facet of RUNX3-dependent tumor suppression. Results TGFb-mediated EMT in non–small cell lung cancer cells The exposure of the non–small cell lung cancer (NSCLC) cell line, A549, to recombinant TGFb (5 ng/mL) induces the acquisition of a mesenchymal morphology in 48 hours (Fig. 1A). TGFb repressed the expression of epithelial markers, CDH1 and OCCLN, Adherens junction, and tight junction proteins, respectively, and increased mesenchymal gene expression of CDH2, FN1, and SERPINE1 (Fig. 1B and C). Western blots displayed fibronectin, vimentin, and N-cadherin upregulation (Fig. 1D, top) and E-cadherin downregulation (Fig. 1D, bottom). The EMT-TFs SNAI1 and SNAI2 were upregulated whereas ZEB1, ZEB2, TWIST1, and TWIST2 remained unaltered (Fig. 1E). SMAD4 depletion abrogated TGFb-mediated EMT, indicating the requirement of canonical SMAD signaling in this context (Supplementary Fig. S1A and S1B). TGFb induces DNA damage in RUNX3-depleted cells To elucidate the function of RUNX3 during TGFb signaling, RUNX3 was silenced using siRNA (RUNX3-KD). Nontargeting oligonucleotides were used as the negative control (CONT-KD; Fig. 1F). Western blot analyses confirmed almost complete depletion of RUNX3 (Fig. 1G). Control and RUNX3-KD cells exhibited similar SMAD2/3 phosphorylation, indicating comparable upstream TGFb signaling (Fig. 1G). RUNX3 levels were downregulated by approximately 40% by TGFb (Fig. 1G, Supplementary Fig. S1C). Interestingly, upon RUNX3 depletion, cells appeared sparser in cell density with abnormal flattened morphology after TGFb exposure (Supplementary Fig. S1D and S1E). Although RUNX3 depletion or TGFb treatment individually reduced cell growth and bromodeoxyuridine (BrdUrd) uptake (80% of control), RUNX3 depletion coupled with TGFb treatment greatly lowered the 24-hour BrdUrd uptake, implying exit from the cell cycle ( , P < 0.001; Supplementary Fig. S1F and S1G). We hypothesized that cells may withdraw from the cell cycle as a stress-dependent response mechanism triggered by DNA damage accumulation. The presence of gH2AX (phosphorylated-histone H2AX at ser139) foci, expressed at DSB sites, was monitored (16). gH2AX antibody was validated using H2AX siRNA (Supplementary Fig. S1H). gH2AX accumulated to a limited extent after TGFb treatment (Fig. 2A). In contrast, approximately 40% of Cancer Res; 78(1) January 1, 2018 89 Krishnan et al. RUNX3-KD cells displayed robust accumulation of gH2AX foci after TGFb treatment. Such gH2AX foci colocalized with 53BP1 and ATM (Ser1981), bona fide markers for DSBs (Fig. 2B–D). DSB accumulation was rescued through the overexpression of siRNAresistant wild-type RUNX3 construct, demonstrating specificity of the knockdown experiments (Supplementary Fig. S2A). Karyotyping confirmed DSB accumulation in approximately 20% of RUNX3-KD cells exposed to TGFb (Supplementary Fig. S2B). To examine the dependence of SMAD signaling on DNA damage accumulation, RUNX3 was codepleted with SMAD4. Following 90 Cancer Res; 78(1) January 1, 2018 TGFb treatment, gH2AX positivity was abolished in RUNX3/ SMAD4 double knockdown (DKD) cells (Fig. 2E and F). In contrast, the codepletion of RUNX3 with SNAI1 and SNAI2 did not impact DNA damage accumulation (Supplementary Fig. S2C and S2D). Thus, an upstream SMAD-dependent, but SNAI1/2independent event, triggered DNA damage following TGFb exposure. The above results were independently validated using another TGFb-responsive NSCLC cell line, NCI-H1299. Similar to A549 cells, TGFb triggered DNA damage upon RUNX3 depletion in NCI-H1299 (Fig. 2G and H). Cancer Research Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Figure 1. Characterization of TGFb-dependent EMT in NSCLC cells. A, A549 cells were seeded into 6-well dishes at the cell density of 0.1  106 cells/well. Cells were either exposed to vehicle control or TGFb. Cells were imaged at the indicated time-points to track spindleshaped morphologic changes induced during TGFb-mediated EMT. Scale bar, 200 mm. B, Gene expression levels of CDH1 and OCLN plotted relative to TGFb-untreated controls. GAPDH was used for normalization of gene expression. C, CDH2, FN1, and SERPINE1 relative gene expression levels are shown. D, Cells were left untreated or exposed to TGFb. Top, Western blots were probed with antibodies against fibronectin, vimentin, and N-cadherin. GAPDH was used as the loading control. Bottom, Western blots were probed with antibodies against E-cadherin. E, Cells were either exposed to vehicle control or TGFb for 48 hours. Relative gene expression levels of SNAI1, SNAI2, TWIST1, TWIST2, ZEB1, and ZEB2 are shown. F, Schematic representation of knockdown experiment. Cells were seeded into 6-cm dishes at the cell density of 0.2  106 cells/well and transfected with pooled siRNA against RUNX3 or control siRNA. On day 4, cells were trypsinized and seeded at the cell density of 0.3  106 cells/6-cm dish and subjected to a second round of siRNA. Twenty-four hours later, either vehicle control or TGFb was added for 48 hours. G, Experiment was done as described in F. Total cell extracts were Western blotted with the indicated antibodies. KU-70 served as the loading control. Graphs show mean  SD. Asterisks represent significant differences.  , P < 0.05;   , P < 0.01;   , P < 0.001; n.s., not significant. DNA Damage during TGFb Signaling Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Figure 2. RUNX3 depletion triggers the accumulation of endogenous DNA damage in the presence of TGFb. A, A549 cells were transfected with control or RUNX3 siRNA and exposed to either vehicle control or TGFb for 48 hours. Samples were fixed and stained with antibody against gH2AX. A zoomed image (200%) of the inset is shown on the right. Scale bar, 50 mm. B, Quantification of gH2AX foci by the "point-maxima method," as described in Materials and Methods. Graph shows percent positivity for gH2AX (left). Right, number of gH2AX foci per cell is shown. Data are representative of four independent experiments. C, Following RUNX3-KD, cells were exposed to TGFb for 48 hours. Top, coimmunofluorescence staining with antibodies against gH2AX and 53BP1. Bottom, coimmunofluorescence staining with antibodies against gH2AX and pATM (Ser1981). Scale bar, 50 mm. D, Experiment was done as described in A. Images were captured using the Arrayscan HCS reader (Cellomics). gH2AX fluorescence intensity per cell was computed and plotted using the Cellomics Arayscan software. E, Cells were subjected to either RUNX3-KD, SMAD4-KD, or RUNX3/SMAD4-DKD and either treated with vehicle control or TGFb for 48 hours. Coimmunofluorescence was performed with gH2AX and 53BP1 antibodies. Images depict immunofluorescence staining for gH2AX and 53BP1 in control cells not exposed to TGFb or in RUNX3-KD, SMAD4-KD, or RUNX3/SMAD4-DKD cells treated with TGFb. Scale bar, 50 mm. F, For experiment described in E, percent cells accumulating >5 gH2AX foci were quantified and plotted. G, NCI-H1299 cells were seeded into 6-cm dishes at the cell density of 0.2  106 cells/dish and transfected with control or RUNX3 siRNA. Following exposure to vehicle control or TGFb for 48 hours, immunofluorescence staining was done using antibody against gH2AX. Scale bar, 40 mm. H, For experiment described in G, relative expression levels of RUNX3 (left) is shown. Right, percent cells expressing greater than 5 gH2AX foci are plotted. Graphs show mean  SD. Asterisks represent significant differences.  , P < 0.05;   , P < 0.01;    , P < 0.001; n.s, not significant. www.aacrjournals.org Cancer Res; 78(1) January 1, 2018 91 Krishnan et al. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Figure 3. Increased intracellular ROS is responsible for heightened DNA damage accumulation in cells undergoing TGFb-mediated EMT. A, A549 cells were subjected to CONT-KD or RUNX3-KD and synchronized in mitosis with nocodazole (100 ng/mL, overnight). Following mitotic shake-off, cells were released into TGFb and BrdUrd (25 mmol/L). After 6 hours, nocodazole (100 ng/mL) was readded to prevent cells from exiting the next cell cycle. Samples were fixed after 24 hours and coimmunofluorescence staining was done using anti-BrdUrd FITC and anti-gH2AX antibodies. Blue arrowheads, gH2AX/BrdUrd double-positive cells; yellow arrowheads, gH2AX–positive/BrdUrd-negative cells. Scale bar, 20 mm. B, RUNX3-KD cells treated with TGFb (48 hours) were fixed for immunofluorescence. Coimmunofluorescence staining was performed with gH2AX antibody and pRPA (Ser33) antibodies. (Continued on the following page.) 92 Cancer Res; 78(1) January 1, 2018 Cancer Research DNA Damage during TGFb Signaling We tested whether the inactivation of other key DNA repair regulators similarly elevated DNA damage after TGFb exposure. Seven candidate genes (ATM, BRCA1, PRKDC, FANCD2, LSD1, PARP1, and TP53) were silenced using siRNA and cells were exposed to TGFb (Supplementary Fig. S2E). Apart from RUNX3 depletion, silencing the above repair genes did not induce DNA damage after TGFb (Supplementary Fig. S2F). RUNX3, by preventing the accumulation of DSBs, therefore performs a unique genome maintenance function following TGFb exposure. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 RUNX3 depletion induced heightened ROS accumulation after TGFb exposure The biological basis underlying the accumulation of DNA damage was investigated. To monitor the cell-cycle phase at which DSBs accumulated, cells were synchronized at mitosis and released into TGFb and BrdUrd-containing media. Coimmunofluorescence revealed that approximately 60% of the cells harboring DNA damage were at G1 (yellow arrowheads; Fig. 3A). In the remaining 40%, gH2AX signals were observed in S-phase cells (blue arrowheads; Fig. 3A) and colocalized with pRPA (Ser33) signals, a protein recruited to the sites of single-stranded DNA (ssDNA) associated with replication stress (Fig. 3B). Endogenous DSBs arise by factors like telomere attrition, impaired chromatin structure, or ROS accumulation (17). Heightened ROS can lead to ssDNA breaks, which at close proximity or upon collision with the transcription or the DNA replication apparatus may convert into DSBs (18). Consistently, the TGFb pathway is known to induce ROS accumulation through two independent pathways—the upregulation of the pro-oxidant, NADPH oxidase, NOX4, and the downregulation of the activity of NRF2, a central antioxidant regulator (19–21). While NOX4 upregulation is elevated by mitochondrial complex III and SMAD signaling, NRF2 inhibition is associated with the elevated expression of its repressive partner ATF3, by TGFb (22). The increased ROS, in turn, promotes redox signaling by regulating TGFb signaling in a feed-forward manner by controlling the activities of the JNK and p38 MAPK pathways (23). We confirmed NOX4 upregulation by TGFb in a SMAD-dependent manner (Supplementary Fig. S3A and S3B). Transcriptional profiling revealed the downregulation of some NRF2-dependent genes like GCLM, GCLC, and GPX2 following TGFb treatment (Supplementary Fig. S3C). While NRF2 expression remained unchanged, ATF3 was transcriptionally upregulated by TGFb (Supplementary Fig. S3D). Pathways known to contribute to increased ROS production during TGFb signaling thus remain conserved in our cellular context. ROS levels were quantified in TGFb-treated cells using carboxyH2DCFDA, a nonfluorescent compound that is converted to a green fluorescent form via ROS-mediated oxidation. Following TGFb exposure, ROS levels increased by approximately 3.8-fold, as indicated by mean fluorescence intensity (MFI) values ( , P < 0.001). However, intracellular ROS levels were elevated by approximately 6-fold in RUNX3-KD cells upon TGFb exposure ( , P < 0.001; Fig. 3C and D). Independently, another reporter reagent CellROX Deep Red Reagent, confirmed that although TGFb exposure and RUNX3-KD independently increased ROS levels to approximately 1.4-fold, the combination of RUNX3 depletion with TGFb exposure increased ROS levels to greater than 2-fold ( , P < 0.01; Fig. 3E and F). Thus, RUNX3 loss results in heightened oxidative stress specifically after TGFb exposure. We asked whether ROS sequestration using antioxidants restores ROS levels to reduce DNA damage. N-acetylcysteine is a pan-antioxidant; deferiprone is an iron chelator while manganese (III) tetrakis (4-Benzoic acid) porphyrin (MnTBAP) is a synthetic metalloporphyrin. Although deferiprone did not rescue gH2AX accumulation (data not shown), both NAC as well as MnTBAP supplementation curtailed the accumulation of gH2AX in RUNX3-KD cells (Fig. 3G and H). The differential effects of these antioxidants on the rescue of DNA damage probably reflect the type of ROS species responsible for DNA damage upon RUNX3 depletion. We conclude that the DNA damage induced by the exposure of RUNX3-deficient cells to TGFb was owing to oxidative stress. HMOX1 downregulation in RUNX3-inactivated cells We reasoned that RUNX3 might regulate the expression of genes involved in combating ROS production after TGFb treatment. Hence, transcriptional profiling was conducted using RNA-Sequencing (RNA-Seq; Fig. 4A). Principal component analysis of the dataset showed the clustering of biological replicates in close proximity, indicating their concordance (Supplementary Fig. S4A). TGFb gene expression signature (126 genes) and EMT scoring methods (254 genes), described earlier were utilized to compare expression signatures (24–26). As expected, TGFb-treated samples showed higher enrichment for TGFb signature (0.14 for control vs. 0.84 after TGFb,  , P < 0.05), validating the Bioinformatic scoring (Fig. 4B). EMT scores were moderately higher upon RUNX3 depletion even in the absence of exogenous TGFb (0.01 vs. 0.29 for CONT-KD and RUNX3-KD, respectively,  , P < 0.05; Fig. 4C). Furthermore, differential gene expression analysis of genes contributing to the EMT score unveiled a subset of genes that underwent downregulation or upregulation after RUNX3 depletion, mimicking TGFb exposure (Supplementary Fig. S4B and S4C). These analyses thus provided a comprehensive genomic perspective on earlier reports of RUNX3 being a negative regulator of EMT (27). Fragments Per Kilobase of transcript per Million mapped reads (FPKM) values of key oxidant and pro-oxidants from the RNA-Seq (Continued.) Blue arrowheads, gH2AX-positive/pRPA-negative; yellow arrowheads, gH2AX/pRPA double-positive cells. Zoomed image of the inset (300%) shows colocalization of gH2AX/pRPA signals. Scale bar, 50 mm. C, Intracellular ROS detection-A549 cells were transfected with either control or RUNX3 siRNA and treated with vehicle control or TGFb for 24 hours and incubated with FITC-labeled carboxy-H2DCFDA reporter. Histogram represents images of carboxy-H2DCFDA fluorescence obtained through flow cytometry. D, For experiment described in C, quantification of MFI per cell, normalized as fold relative to CONT-KD cells not treated with TGFb. E, A549 cells were transfected with either control or RUNX3 siRNA and treated with vehicle control or TGFb for 24 hours Cells were incubated with APC-labeled Cell-ROX reporter reagent. Histogram represents images of Cell-ROX fluorescence obtained through flow cytometry. F, For experiment described in E, quantification of MFI per cell normalized as fold relative to CONT-KD cells not treated with TGFb. G, Experiment was carried out as described in the schematic. N-acetylcysteine (5 mmol/L) or MnTBAP chloride (100 mmol/L) was added 24 hours prior to TGFb exposure for 48 hours. Immunofluorescence staining was performed with gH2AX antibody. Scale bar, 50 mm. H, Percent cells containing greater than 5 gH2AX foci per cell were counted and are represented in the graph (n ¼ 300). Two independent experiments were done as triplicates. Graphs show mean  SD. Asterisks represent significant differences.  , P < 0.05;   , P < 0.01;    , P < 0.001; n.s, not significant. www.aacrjournals.org Cancer Res; 78(1) January 1, 2018 93 Krishnan et al. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Figure 4. HMOX1 downregulation is responsible for increased DNA damage in RUNX3-depleted cells undergoing TGFb-mediated EMT. A, A549 cells were seeded into 6-cm dishes at the cell density of 0.2  106 cells/dish and transfected with control siRNA or RUNX3 siRNA. Three days later, cells were trypsinized and seeded at the cell density of 0.3  106 cells/6-cm dish and subjected to a second round of siRNA. Twenty-four hours later, cells were exposed to either vehicle control or TGFb for 48 hours. Cells were harvested and RNA extraction and sequencing were performed. RNA-Seq reads were mapped to human genome HG19 and FPKM values were calculated per gene. B, Graph depicts the average TGFb enrichment score calculated across biological replicates. C, Graph depicts the average EMT score computed for biological replicates. D, Heatmap depicting expression of genes encoding for antioxidants. Gene expression derived from FPKM values was plotted. Biological replicate for each sample is represented individually as 1 and 2. (Continued on the following page.) 94 Cancer Res; 78(1) January 1, 2018 Cancer Research DNA Damage during TGFb Signaling robustly upregulated within 6 hours after H2O2 exposure, the knockdown of RUNX3 markedly attenuated HMOX1 induction (Supplementary Fig. S5D). The data presented above using several cancer lines and genomic datasets thus establish RUNX3 as an important regulator of HMOX1 transcription. Ectopic HMOX1 expression prevents TGFb-elicited DNA damage in RUNX3-depleted cells The consequence of HMOX1 downregulation to DNA damage accumulation was then investigated. GFP-tagged HMOX1 was introduced ectopically in RUNX3-depleted cells, following which, cells were treated with TGFb. As a positive control, a siRNAresistant version of wild-type GFP-tagged RUNX3 was overexpressed. ROS levels were compared in the GFP-positive and GFPnegative populations using the CellROX assay. Cellular ROS amounts restored to basal levels upon RUNX3 overexpression, indicating the specificity of the siRNA experiments ( , P < 0.01; Fig. 4J). Importantly, HMOX1 overexpression reduced ROS levels to basal levels upon RUNX3 depletion ( , P < 0.01; Fig. 4J). Next, gH2AX positivity was scored following HMOX1 overexpression and TGFb exposure. As a positive control, a siRNA-resistant version of wild-type GFP-tagged RUNX3 was overexpressed. The DNA-binding deficient mutant of RUNX3 (R178Q) was also tested for its ability to rescue DNA damage. Importantly, RUNX3-KD cells overexpressing HMOX1 or wild-type (WT)RUNX3 did not accumulate DNA damage after TGFb exposure ( , P < 0.001; Fig. 4K and L; Supplementary Fig. S6A and S6B). Surprisingly, R178Q-RUNX3 mutant also rescued DNA damage in the presence of TGFb (Supplementary Fig. S6B). We conclude HMOX1 downregulation as the mechanism underlying increased DNA damage accumulation in RUNX3-deficient cells. Intriguingly, the genome maintenance function of RUNX3 does not seem to rely on its DNA-binding property (elaborated further in "Discussion"). Many intracellular signaling cascades and transcription factors have been found to regulate HMOX1 transcription (31). Of these, HMOX1 transcription is mainly regulated by the master regulator of antioxidant gene expression, NRF2 (32). HMOX1 enhancer is bound by the BACH1 repressor complex at the basal transcriptional state, but upon oxidative stress, BACH1 is replaced by NRF2, to stimulate HMOX1 transcription (33). Taking advantage of the BACH1-dependent repression of HMOX1, HMOX1 levels were increased in RUNX3-KD cells by silencing BACH1 (Supplementary Fig. S6C and S6D). BACH1/RUNX3-DKD cells did not accumulate DNA damage upon TGFb exposure ( , P < 0.001; Supplementary Fig. S6E and S6F), indicating that DNA damage accumulation can be rescued by genetic manipulations that elevate HMOX1 levels. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 datasets were assessed. Pro-oxidant gene expression largely remained unchanged upon RUNX3 depletion (data not shown). In contrast, among the 21 critical antioxidant defense genes examined, RUNX3 depletion resulted in transcriptional downregulation of the metabolic enzyme and a redox regulator, Heme oxygenase-1 (HMOX1 or HO-1; Fig. 4D and E). HMOX1 is a ratelimiting enzyme of an important metabolic pathway that detoxifies heme (Fe-protoporphyrin IX), an important cofactor of several Heme-containing proteins like hemoglobin, cytochrome c, and catalase. Three metabolic byproducts are generated by HMOX1-carbon monoxide, heme and biliverdin, of which, biliverdin is reduced by the enzyme biliverdin reductase to generate the antioxidant, bilirubin (28). The downregulation of HMOX1 following RUNX3 depletion was validated by qPCR in gastric (AZ521, HGC-27) and lung cancer cells (NCI-H23, PC-14, NCI-H1299), two tissues where RUNX3 is an established tumor suppressor. RUNX3 knockdown attenuated HMOX1 expression by approximately 50%–80% in every cell line examined (Fig. 4F). HMOX1 expression levels also strongly correlated with RUNX3 status in Spearman correlation analysis of gene expression of 436 samples in TCGA (The Cancer Genome Atlas Consortium) lung adenocarcinoma datasets (Supplementary Fig. S5A). In hierarchical clustering analysis, lung cancers with high RUNX3 levels had either high HMOX1 or intermediate HMOX1 expression, while lung cancers with low RUNX3 levels had low HMOX1 expression (r ¼ þ0.3028; P ¼ 1.33e10). Western blot analysis confirmed that basal and TGFbinduced HMOX1 protein levels were downregulated after RUNX3 depletion (Fig. 4G). According to earlier reports, TGFb addition induces HMOX1 in human retinal pigment epithelial cells, A549 cells, and bovine choroid fibroblasts (29). In contrast, TGFb suppressed HMOX1 expression through the elevated expression of the negative regulators, MafK and Bach1, in NMuMG (30). In our analysis, HMOX1 was transcriptionally upregulated by TGFb across three different cell lines, A549, NCI-H1299, and NCI-H292 (Supplementary Fig. S5B). The induction of MMP-9 by TGFb was used as a marker for TGFb signaling. We then examined the effect of RUNX3 status on HMOX1 induction by TGFb. Upon RUNX3 depletion, HMOX1 transcript and protein levels were substantially diminished across all timepoints examined (maximal expression at 8 hours following TGFb exposure; Fig. 4H and I; Supplementary Fig. S5C). Thus, RUNX3-dependent HMOX1 accumulation is an early response following TGFb exposure. Finally, we evaluated whether RUNX3-mediated HMOX1 upregulation was modulated by ROS production. Cells were challenged with acute oxidative stress (250 mmol/L hydrogen peroxide) and HMOX1 levels were measured. Whereas HMOX1 was (Continued.) E, A549 cells were subjected to control or RUNX3 knockdown for 72 hours. HMOX1 levels were measured using qPCR. F, Cell lines shown in the graph were transfected with RUNX3 for 72 hours. HMOX1 levels were measured using qPCR. G, Cells were transfected with either CONT siRNA or RUNX3 siRNA and treated with vehicle control or TGF-b for 36 hours. Samples were Western blotted with RUNX3 and HMOX1-specific antibodies. GAPDH was used as the loading control. H, Experiment was performed as described in A. Cells were exposed to TGFb and harvested at the indicated timepoints, followed by qPCR analysis of HMOX1. I, CONT-KD or RUNX3-KD cells were exposed to TGFb and harvested at the indicated timepoints. Western blot analysis of RUNX3 and HMOX1 is shown. GAPDH was used as the loading control. J, Cells were transfected with RUNX3 siRNA. GFP-tagged siRNA-resistant RUNX3 or GFP-HMOX1 was expressed for 24 hours, following which, cells were exposed to TGFb for another 24 hours. Cells were incubated with Cell-ROX reagent and subjected to flow cytometry. The MFI for Cell-ROX fluorescence in the GFP-positive and GFP-negative populations are shown. K, Experiment was performed as described in A. GFP-HMOX1 was expressed 24 hours before TGFb addition. Following 48 hours of TGFb exposure, cells were fixed and stained with 53BP1 antibody. Scale bar, 50 mm. L, For experiment described in K, percent cells expressing greater than five 53BP1 foci/cell were quantified within the GFP-positive and GFP-negative populations (n ¼ 300). Two independent experiments were done as triplicates. For graphs shown in B and C, data represent mean  SEM. For all other graphs, data represent mean  SD. Asterisks represent significant differences.  , P < 0.05;  , P < 0.01;   , P < 0.001; n.s, not significant. www.aacrjournals.org Cancer Res; 78(1) January 1, 2018 95 Krishnan et al. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Figure 5. RUNX3 depletion induces cellular senescence and inflammatory cytokine expression in cells undergoing TGFb-mediated EMT. A, Cells were transfected with control or RUNX3 siRNA and treated with vehicle control or TGFb for 48 hours. Immunofluorescence staining was performed with gH2AX antibody and Alexa Fluor-594–conjugated phalloidin. The phalloidin and gH2AX images were imaged individually and merged using Adobe Photoshop software. Merged images represent coimmunofluorescence for phalloidin and gH2AX staining. Scale bar, 50 mm. B, Experiment was performed as described in A. Scatter plot shows projected nuclear area across samples. C, Experiment was performed as described in A. Images were captured and subjected to quantitative image analysis to compute nuclear area. The average gH2AX foci/cell within cells with the nuclear area 100–250 mm2 or greater than 250 mm2 is shown. (Continued on the following page.) 96 Cancer Res; 78(1) January 1, 2018 Cancer Research DNA Damage during TGFb Signaling TGFb induces SASP in RUNX3-depleted cells Although senescence was originally discovered as a tumorsuppressive mechanism, it is now well established that the senescent response can be also protumorigenic. In a phenomenon referred to as the SASP and regulated by factors like NFkB, C/EBPb, mTOR, MacroH2A.1, and GATA4, senescent cells secrete a large repertoire of cytokines and metalloproteinases (35). The sustenance of senescent cells within tumors thus creates a pro-oncogenic inflammatory microenvironment promoting tumor progression (36). Genes differentially expressed were subjected to NOISeq filtration (methods) and GO analysis (Supplementary Table S1). Interestingly, the terms "cytokine-activity", "metalloproteinases" were the "molecular functions" enriched in the RUNX3-depleted transcriptomes when compared with control cells exposed to TGFb (Fig. 5G). The cluster of genes representing this group CXCL1, INHBA, BMP2, CCL2, CXCL3, CXCL2, IL32, IL8, AREG, GDF15, and IL1A represent cytokines and chemokines secreted by senescent cells exhibiting the SASP phenotype (36). The metalloproteinase gene cluster encoding MMP10, MMP9, MMP2, and MMP1 are also enzymes known to be enriched in the secretomes of senescent cells (Fig. 5H, heatmap). Consistently, miR-146a, a miRNA induced during DNA damage-induced senescence and SASP, was highly expressed only when RUNX3-depleted cells were exposed to TGFb (37) (Fig. 5I). The differential upregulation of SASP genes in RUNX3-KD cells exposed to TGFb was confirmed through qPCR analysis (Fig. 5J). Cellular senescence is ATM- and ATR-dependent but p53independent In cells harboring unrepaired DNA damage, the DDR kinases ATM and ATR relieve the inhibition of p53 and p16INK4a to induce growth arrest and senescence. To assess the roles of ATM, ATR, and p53 in the induction of senescence, we inhibited ATM and ATR using two specific small-molecule inhibitors, KU-55933 (10 mmol/L) and VE-821 (2 mmol/L), respectively (38, 39). To study the function of p53 as a senescence inducer, p53 was codepleted with RUNX3 using siRNA. The role of p16INK4a was not studied as this gene is deleted in A549 cells. To confirm the inactivation of the ATM/ATR pathways by the small-molecule inhibitors, immunofluorescence staining was performed with pATM/ATR substrate antibody that detects ATM and ATR substrates phosphorylated at the consensus SQ/TQ sites. The concurrent inhibition of ATM and ATR significantly reduced both pSQ/TQ signals and gH2AX positivity (Fig. 6A). Importantly, the inactivation of ATM and ATR significantly mitigated cellular senescence, establishing a role for these DDR kinases as transducers of senescence signals ( , P < 0.001; Fig. 6B and C). In contrast, p53 depletion had little effect on alleviating the extent of senescence (Fig. 6B and C). Taken together, the senescence activated in RUNX3-deficient cells upon TGFb exposure was reliant on upstream DNA damage signaling by ATM and ATR but was independent of p53 status. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 TGFb provokes cellular senescence in RUNX3-depleted cells after TGFb exposure The accumulation of DNA damage either results in apoptosis or persistently activates the DNA damage checkpoint to mediate cellular senescence (34). As increased apoptosis was not observed in RUNX3-KD cells exposed to TGFb, we asked whether cells underwent senescence. Senescence is a cell-fate associated with flattened cell shape, higher nuclear area, cell-cycle exit, and positivity for the SA-b-gal. Indeed, RUNX3-depleted cells upon TGFb exposure were more flattened in cell shape and exhibited atypical disorganized actin, as evident by staining with the F-actin probe, Phalloidin (Fig. 5A). RUNX3-KD cells upon TGFb exposure also had a significantly larger nuclear area ( , P < 0.001; Fig. 5B). Image quantification revealed a close correlation between nuclear area and gH2AX focus accumulation (r ¼ 0.382; Fig. 5C). Finally, approximately 40%–50% of RUNX3-KD cells exposed to TGFb exhibited positivity for the senescence hallmark, SA-b-gal (Fig. 5D and E). Thus, whereas TGFb triggers EMT in cells with intact RUNX3, DNA damage and senescence are triggered upon the loss of RUNX3. To study whether reduced HMOX1 was responsible for senescence following RUNX3 depletion, rescue experiments were performed. GFP-HMOX1 was overexpressed and cells were subjected to flow cytometry to sort for GFP-positive cells. As a positive control, siRNA-resistant wild-type GFP-RUNX3 was overexpressed. GFP-positive cells were cultured and exposed to TGFb for 48 hours. Cellular senescence was substantially rescued in RUNX3-depleted cells upon the ectopic expression of wild-type RUNX3 or HMOX1 (Fig. 5F). Taken together, the instatement of HMOX1 expression in RUNX3-depleted cells restored cellular ROS, rescued DNA damage, and premature senescence following TGFb exposure. TGFb provokes DNA damage and senescence upon RUNX1 deficiency We asked whether RUNX1 and RUNX2 perform redundant roles with RUNX3 in genome maintenance during TGFb signaling. qPCR analysis confirmed efficient siRNA-mediated depletion of RUNX1, RUNX2, and RUNX3 (Supplementary Fig. S7A). Interestingly, TGFb exposure triggered the accumulation (Continued.) One-way ANOVA nonparametric test was used for statistical analysis.    , P < 0.001. D, Cells were treated as described in A. Samples were fixed and stained for SA-b-gal enzyme overnight at 37
C. Inset image was zoomed (200%) and is shown below. Scale bar, 50 mm. E, SA-b-gal–positive cells (blue) were scored and plotted. Quantification of SA-b-gal positivity across two independent experiments, each done as triplicates. F, Cells were transfected with RUNX3 siRNA. GFP-tagged siRNA-resistant RUNX3 or GFP-HMOX1 were expressed for 24 hours, following which, cells were sorted by flow cytometry into GFP-positive and GFP-negative populations and plated. After TGFb exposure for 48 hours, the SA-b-gal assay was performed. G, RNA-Seq data described in Fig. 4A was subjected to pathway analysis. Gene ontology (GO) pathways enriched under the category "Molecular function" are shown. Control and RUNX3-KD cells treated with TGFb were subjected to differential gene expression analysis, as described in Materials and Methods. Genes regulated by RUNX3 in a TGFb-independent manner were excluded. The online tool, Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.7, was used for GO analysis. Genes expressing more than 200% of control were considered upregulated, whereas those with expression less than 50% of control were considered as downregulated. List of differentially expressed genes are shown in Supplementary Table S1. H, Heatmap depicting overexpression of SASP genes. Gene expression derived from FPKM values was plotted. Biological replicates for each sample, is represented individually. I, miR-146a FPKM values derived from RNA-Seq analyses are shown. J, Experiment was performed as described in A. Samples were harvested after 48 hours of TGFb treatment. qPCR analyses were done for the indicated genes. Graphs show mean  SD. Asterisks represent significant differences.  , P < 0.05;   , P < 0.01;    , P < 0.001; n.s, not significant. www.aacrjournals.org Cancer Res; 78(1) January 1, 2018 97 Krishnan et al. of DNA damage following both RUNX1 and RUNX3 depletion but not following RUNX2 depletion (Supplementary Fig. S7B and S7C). The combined knockdown of RUNX1 and RUNX3 did not further increase the extent of DNA damage accumulation (RUNX1-KD vs. RUNX3-KD vs. RUNX1/RUNX3-DKD20% 35%, and 38% gH2AX positivity, respectively). RUNX1 and RUNX3 depletion significantly reduced HMOX1 levels after TGFb exposure and their combined knockdown did not further exacerbate the extent of HMOX1 downregulation (Supplementary Fig. S7D). Finally, RUNX1 depletion also triggered senescence following TGFb treatment although to a lesser extent than that following RUNX3 depletion (Supplementary Fig. S7E). Taken together, although both RUNX1 and RUNX3 are required to prevent the accumulation of DNA damage following TGFb exposure, RUNX3 depletion evoked stronger DNA damage phenotypes and HMOX1 downregulation than RUNX1. HMOX1 depletion radiosensitizes cells Although increased ROS levels promote tumorigenesis, it has been proposed that escalating ROS to lethal levels has the potential to block tumor progression by inducing apoptosis. Moreover, increased ROS levels can accentuate the cytotoxicity of some chemotherapeutics like cisplatin, anthracyclines, and irradiation (41). Indeed, HMOX1 inhibition through its competitive smallmolecule inhibitor, Zinc protoporphyrin IX synergizes with cisplatin (42). 98 Cancer Res; 78(1) January 1, 2018 Discussion Here, we uncovered that the pro-oncogenic effects for TGFb signaling are exacerbated upon the loss of RUNX3. We show that the reduced expression of HMOX1 in RUNX3-deficient cells caused a catastrophic imbalance toward greater ROS production by TGFb, causing oxidative DNA damage and genomic instability. Consequent to DNA damage, the DDR kinases ATM and ATR mediate senescence, which is accompanied by the expression of proinflammatory cytokines. The latter phenomenon, known as SASP, is known to favor tumor progression (Fig. 7H). Importantly, the inactivation of major DNA repair genes in TGFb-stimulated cells did not result in DNA damage, indicating a unique role for RUNX3 in the defense against TGFb-dependent promotion of cancer progression. Over the years, we and others have studied the tumor suppressing potential of RUNX3 in early-stage cancers (43). The inactivation of RUNX3 at the earliest stages of cancer progression seems to be a mechanism evolved by cancer cells to simultaneously suppress the antiproliferative arm of TGFb signaling while accentuating its pro-oncogenic roles. In support, RUNX3 loss impairs p21 and Bim transcription during TGFb signaling, thus impairing apoptosis while promoting gastric hyperplasia (13, 14). Second, RUNX3 loss supports pro-oncogenic TGF-b signaling by spontaneously upregulating a subset of EMT genes, thus mimicking a partial EMT phenotype (Supplementary Fig. S4B and S4C; Fig. 4C; ref. 27). Finally, our work here introduces another independent and novel aspect of the TGFb–RUNX3 axis wherein RUNX3 loss promotes further ROS accumulation during TGFb signaling, thus triggering genomic instability. In support, genomic analyses of TCGA lung adenocarcinoma datasets showed that mutation load and copy number alterations of tumors harboring a high-TGFb signature increased following RUNX3 inactivation. The observation that RUNX3 levels reduce following TGFb exposure might seem at odds with the genome protection role for RUNX3 identified in this study. It is noteworthy that despite the reduction in RUNX3 levels, TGFb-treated cells do not accumulate DNA damage, implying that the lower levels of RUNX3 are Cancer Research Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Exacerbation of genomic instability in RUNX3-inactivated lung cancers with "a TGFb high" expression gene signature To assess the in vivo significance of RUNX3 loss specifically in the context of TGFb signaling, we interrogated TCGA datasets of lung adenocarcinomas and evaluated the relationship between genomic instability, RUNX3 status, and TGFb signature. In human lung adenocarcinomas and mouse models, RUNX3 shows tumor-suppressive function and its inactivation promotes lung cancer progression (40). Genomic instability was measured as mutation rate (number of mutations accumulation per mega-base of genome) and copy number alteration or CNA (presence of genes with copy number 3 or 1). First, tumors were stratified as either "TGFb-high" or "TGFb-low" based on TGFb expression signature (Fig. 7A and B). Then, such tumors were further stratified as "RUNX3-high" or "RUNX3-low" (median gene expression values). No significant differences in genomic instability were found when "TGFb-high" and "TGFb-low" tumors were compared or when "RUNX3-low" and "RUNX3-high" tumors were compared. In contrast, when "TGFb-high" tumors were further stratified on the basis of RUNX3 status, tumors with low RUNX3 levels had significantly greater CNA (P ¼ 0.0001) and accumulation of mutations (P ¼ 0.0221) as compared with tumors with high RUNX3 expression (Fig. 7A and B). Interestingly, the tumors that fell into the quadrant of "TGFb high/RUNX3 low" exhibited a trend towards poorer survival as compared with those tumors that fell into the quadrant of "TGFb low/RUNX3 high" (P ¼ 0.095, HR ¼ 1.845, n ¼ 181; Fig. 7C). A comparison of overall survival with RUNX3 and TGFb status are shown in Supplementary Fig. S8A and S8B. Thus, closely recapitulating the data shown in earlier sections, RUNX3 inactivation in lung adenocarcinomas exacerbates genomic instability and survival of tumors harboring a TGFb expression signature. We explored the possibility of exploiting increased oxidative stress as a therapeutic strategy against TGFb-high/RUNX3-low tumors. Cells were exposed to increasing doses of radiation and percent survival was plotted. TGFb treatment of RUNX3-depleted cells resulted in a 5-fold increase in radiosensitization (Fig. 7D and E). Finally, we depleted HMOX1 using siRNA and studied whether it recapitulates the loss of RUNX3 in enabling radiosensitization following TGFb. Similar to RUNX3 loss, siRNA-mediated HMOX1 knockdown resulted in high levels of oxidative stress, specifically upon exposure to TGFb ( , P < 0.001; Supplementary Fig. S8C–S8E). HMOX1 depletion also engendered cellular senescence and DNA damage, although gH2AX signals generated were more diffuse than that following RUNX3 depletion (discussed below; Supplementary Fig. S8F–S8H). The depletion of HMOX1 also rendered radiosensitization similar to RUNX3 depletion. More importantly, the extent of radiosensitization mediated by HMOX1 depletion was increased by TGFb (Fig. 7F and G). Put together, the above data emphasize oxidative stress induction through HMOX1 inhibition as a therapeutic vulnerability for tumors exposed to pro-oncogenic TGFb. DNA Damage during TGFb Signaling Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Figure 6. RUNX3 depletion induces cellular senescence in an ATM/ATR dependent, but p53-independent manner. A, A549 cells were subjected to CONT-KD, RUNX3-KD, or RUNX3/p53-DKD for 3 days. As indicated, cells were pretreated with either vehicle (DMSO), KU-55933 (10 mmol/L), or VE-821 (2 mmol/L) for 2 hours, following which, they were exposed to TGFb for 48 hours in the presence of the small-molecule inhibitors. Coimmunofluorescence staining was performed with antibodies recognizing pSQ/TQ sites and gH2AX. Zoomed (200%) image of the inset is shown on the right. Percent cells expressing greater than 5 gH2AX foci are indicated within the parentheses. Scale bar, 50 mm. B, For the experiment described in A above, cells were fixed and stained for SA-b-gal enzyme overnight at 37
C. Scale bar, 100 mm. C, SA-b-gal–positive cells (blue) were scored and plotted. Graphs show mean  SD. Asterisks represent significant differences.  , P < 0.05;   , P < 0.01;    , P < 0.001; n.s, not significant. www.aacrjournals.org Cancer Res; 78(1) January 1, 2018 99 Krishnan et al. Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 Figure 7. Exacerbation of genomic instability and radiosensitization upon RUNX3 or HMOX1 depletion in the presence of TGFb. A, Lung adenocarcinomas were stratified on the basis of TGFb signature and RUNX3 expression levels utilizing median expression values. Copy number alterations are shown in the scatter plot with bar (mean  SEM). B, Lung adenocarcinomas were stratified on the basis of TGFb signature and RUNX3 expression levels utilizing median expression values. Mutations per megabase of the genome are shown in the scatter plot with bar (mean  SEM). C, Kaplan–Meier survival curve of the indicated cohorts is shown. n ¼ 181, P ¼ 0.0955, HR ¼ 1.845. Statistical analysis was done with the Mann–Whitney U test. D, A549 cells were subjected to control or RUNX3-KD and exposed to vehicle control or TGFb for 48 hours. Cells were exposed to increasing doses of X-irradiation in clonogenic assay. E, For experiment described in D, percent survival was plotted relative to radiation-untreated controls. F, A549 cells were subjected to control or HMOX1-KD and exposed to TGFb or vehicle control. Forty-eight hours following TGFb addition, cells were exposed to increasing doses of X-irradiation. G, For experiment described in F, percent survival was plotted relative to radiation-untreated controls. H, RUNX3 provides defense against TGFb-dependent promotion of cancer progression. Top, TGFb signaling induces ROS production, which participates in redox signaling and promotes TGFb-mediated EMT. However, RUNX3 mitigates the accumulation of excessive ROS via the transcriptional upregulation of the HMOX1 and thus curtails tumor progression. Under these conditions, where RUNX3 levels are intact, TGFb-mediated EMT is associated with low levels of DNA damage and genomic instability. Bottom, in the absence of RUNX3, low HMOX1 levels perpetrate high levels of ROS, which promotes cancer progression by the induction of DNA damage and senescence. Cellular senescence, in turn, fuels the production of inflammatory cytokines that further amplify the pro-oncogenic state. The reinstatement of HMOX1 expression in RUNX3-depleted cells restored cellular ROS, rescued DNA damage and premature senescence in the presence of TGFb. Thus, HMOX1 downregulation seems to be one of the major mechanisms by which genomic instability is brought about by the TGFb exposure of RUNX3-deficient cells. 100 Cancer Res; 78(1) January 1, 2018 Cancer Research DNA Damage during TGFb Signaling triggering high levels of ROS through HMOX1 inhibition can promote greater radiosensitivity following TGFb exposure. In summary, our studies imply intriguing collaboration between an intrinsically inactivated tumor suppressor, RUNX3, and an extrinsically acting cytokine, TGFb in the promotion of DNA damage and genomic instability. Dissecting how genomic instability is exacerbated through a crosstalk between cancer cell– autonomous and nonautonomous factors is likely to create new therapeutic targets and immune-related approaches to tackle TGFb-driven tumors. Disclosure of Potential Conflicts of Interest J.P. Thiery is a visiting professor at the University Of Bergen and is a consultant/advisory board member for Aim Biotech Singapore and Actgenomics Taipei. No potential conflicts of interest were disclosed by the other authors. Authors' Contributions Conception and design: V. Krishnan, M.D. Kulkarni, J.P. Thiery, Y. Ito Development of methodology: V. Krishnan, M.D. Kulkarni, L.S. Tay, A. Ganesan Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Krishnan, Y.L. Chong, M.D. Kulkarni, M.B. Bin Rahmat, D.S. Jokhun Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Krishnan, T.Z. Tan, M.D. Kulkarni, D.S. Jokhun, D. Chih-Cheng Voon, GV Shivashankar, J.P. Thiery Writing, review, and/or revision of the manuscript: V. Krishnan, Y.L. Chong, M.D. Kulkarni, L.S.H. Chuang, Y. Ito Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Study supervision: Y. Ito Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 competent in conferring genome protection. Interestingly, a recent seminal study has showed that expression levels of Runx3 respond to and combine with Smad4 status to regulate the balance between cell division and dissemination in pancreatic cancers (15). Indeed, data mining of ChIP-sequencing datasets reveal that the RUNX3 promoter harbors elements that bind to TFs downstream of TGFb pathway like SMAD4, SOX2, FOXO1, FOXC1, and EZH2 (44). We speculate that the binding of one of these TFs to the RUNX3 promoter might represent a feedback mechanism to alter RUNX3 levels during TGFb signaling In previous studies, ROS levels were shown to elevate between 8 and 16 hours following TGFb exposure and subside to basal levels within 48 hours, suggesting temporal regulation of ROS production (20, 45). Our observations that HMOX1 is transcribed in a RUNX3-dependent manner within 8 hours, suggests that cells have evolved this mechanism to mitigate the accumulation of tumor-promoting ROS during TGF-b pathway. It is currently unclear if HMOX1 induction by RUNX3 is via the direct binding of RUNX3 to the HMOX1 promoter or through indirect means. Interestingly, genome-wide RUNX3-CHIP data have revealed the binding of RUNX3 to a super enhancer region located -3.5 kb upstream of the HMOX1 promoter (Encode consortium). However, the indirect regulation of HMOX1 transcription by RUNX3 is also plausible, because the DNA binding mutant, R178-RUNX3 was also capable of rescuing DNA damage accumulation. In earlier studies, it has been shown that RUNX2 binds and recruits CEBPd to promoter enabling transcription and precluding the requirement for direct DNA binding (46). Alternatively, since RUNX3 is known to establish multiple protein-interacting partners, it is possible that RUNX3 might control HMOX1 transcription indirectly through interaction with transcription factors known to regulate HMOX1 levels (47). Aside from being a major regulator of HMOX1 expression, we speculate that RUNX3 has additional roles in genome maintenance upon TGFb exposure. This is based on the observation that the gH2AX foci generated by RUNX3 loss were higher in number and more distinct as compared to the more diffuse and lesser gH2AX signals generated by HMOX1 knockdown. One possible explanation is that HMOX1 depletion creates oxidative stress and single-stranded breaks that may not be efficiently converted in DSBs. Given that RUNX proteins have added roles in DNA repair, we speculate that the absence of RUNX3 might additionally disrupt repair of oxidative stress–mediated DNA damage (48, 49). Our work raises novel possibilities of exploiting ROS-dependent genomic instability for the therapeutic targeting of cancers exposed to pro-oncogenic TGFb. This is based on the rationale that increasing ROS to very high levels has the potential to block tumor progression by inducing cancer cell senescence or apoptosis (41, 50). Accordingly, our studies (Fig. 7F and G) suggest that Acknowledgments The authors would like to thank Ms. Shu Ying from the Confocal facility, National University of Singapore, for help with image analysis and the Cytogenetic Facility of the Genome Institute of Singapore for Karyotyping services. We would like to thank Ms. Charell Lim, Ms. Charmaine Nai and Ms. Soundharya Ravindran for their contributions during the initial phases of this work. The research is supported by the New Investigator Grant (CBRG-NIG; provided to V. Krishnan) from the National Medical Research Council (NMRC), Singapore (grant number-NMRC/BNIG/2024/2014) and by the National Research Foundation (NRF) and the Singapore Ministry of Education under its Research Centres of Excellence initiative and by the NRF under its Translational and Clinical Research Flagship Programme (provided to Y. Ito; grant numberNMRC/TCR/009-NUHS/2013). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Received April 25, 2017; revised September 11, 2017; accepted October 23, 2017; published OnlineFirst October 26, 2017. References 1. Negrini S, Gorgoulis VG, Halazonetis TD. Genomic instability–an evolving hallmark of cancer. Nat Rev Mol Cell Biol 2010;11:220–8. 2. Tubbs A, Nussenzweig A. Endogenous DNA damage as a source of genomic instability in cancer. Cell 2017;168:644–56. 3. Massague J. TGFbeta in cancer. Cell 2008;134:215–30. 4. David CJ, Huang YH, Chen M, Su J, Zou Y, Bardeesy N, et al. TGF-beta tumor suppression through a lethal EMT. Cell 2016;164:1015–30. 5. Hasegawa Y, Takanashi S, Kanehira Y, Tsushima T, Imai T, Okumura K. Transforming growth factor-beta1 level correlates with angiogenesis, www.aacrjournals.org tumor progression, and prognosis in patients with nonsmall cell lung carcinoma. Cancer 2001;91:964–71. 6. Yu JR, Tai Y, Jin Y, Hammell MC, Wilkinson JE, Roe JS, et al. TGF-beta/Smad signaling through DOCK4 facilitates lung adenocarcinoma metastasis. Genes Dev 2015;29:250–61. 7. Thiery JP, Acloque H, Huang RY, Nieto MA. Epithelial-mesenchymal transitions in development and disease. Cell 2009;139:871–90. 8. Moustakas A, Heldin CH. Induction of epithelial-mesenchymal transition by transforming growth factor beta. Semin Cancer Biol 2012;22:446–54. Cancer Res; 78(1) January 1, 2018 101 Krishnan et al. 102 Cancer Res; 78(1) January 1, 2018 28. Stocker R, Yamamoto Y, McDonagh AF, Glazer AN, Ames BN. Bilirubin is an antioxidant of possible physiological importance. Science 1987;235: 1043–6. 29. Zarjou A, Agarwal A. Heme oxygenase-1 as a target for TGF-beta in kidney disease. Semin Nephrol 2012;32:277–86. 30. Okita Y, Kamoshida A, Suzuki H, Itoh K, Motohashi H, Igarashi K, et al. Transforming growth factor-beta induces transcription factors MafK and Bach1 to suppress expression of the heme oxygenase-1 gene. J Biol Chem 2013;288:20658–67. 31. Lavrovsky Y, Schwartzman ML, Levere RD, Kappas A, Abraham NG. Identification of binding sites for transcription factors NF-kappa B and AP-2 in the promoter region of the human heme oxygenase 1 gene. Proc Natl Acad Sci U S A 1994;91:5987–91. 32. Alam J, Stewart D, Touchard C, Boinapally S, Choi AM, Cook JL. Nrf2, a Cap'n'Collar transcription factor, regulates induction of the heme oxygenase-1 gene. J Biol Chem 1999;274:26071–8. 33. Igarashi K, Sun J. The heme-Bach1 pathway in the regulation of oxidative stress response and erythroid differentiation. Antioxid Redox Signal 2006;8:107–18. 34. Campisi J, d'Adda di Fagagna F. Cellular senescence: when bad things happen to good cells. Nat Rev Mol Cell Biol 2007;8:729–40. 35. Davalos AR, Coppe JP, Campisi J, Desprez PY. Senescent cells as a source of inflammatory factors for tumor progression. Cancer Metastasis Rev 2010;29:273–83. 36. Acosta JC, O'Loghlen A, Banito A, Guijarro MV, Augert A, Raguz S, et al. Chemokine signaling via the CXCR2 receptor reinforces senescence. Cell 2008;133:1006–18. 37. Kang C, Xu Q, Martin TD, Li MZ, Demaria M, Aron L, et al. The DNA damage response induces inflammation and senescence by inhibiting autophagy of GATA4. Science 2015;349:aaa5612. 38. Hickson I, Zhao Y, Richardson CJ, Green SJ, Martin NM, Orr AI, et al. Identification and characterization of a novel and specific inhibitor of the ataxia-telangiectasia mutated kinase ATM. Cancer Res 2004;64:9152–9. 39. Reaper PM, Griffiths MR, Long JM, Charrier JD, Maccormick S, Charlton PA, et al. Selective killing of ATM- or p53-deficient cancer cells through inhibition of ATR. Nat Chem Biol 2011;7:428–30. 40. Lee YS, Lee JW, Jang JW, Chi XZ, Kim JH, Li YH, et al. Runx3 inactivation is a crucial early event in the development of lung adenocarcinoma. Cancer Cell 2013;24:603–16. 41. Gorrini C, Harris IS, Mak TW. Modulation of oxidative stress as an anticancer strategy. Nat Rev Drug Discov 2013;12:931–47. 42. Fang J, Sawa T, Akaike T, Greish K, Maeda H. Enhancement of chemotherapeutic response of tumor cells by a heme oxygenase inhibitor, pegylated zinc protoporphyrin. Int J Cancer 2004;109:1–8. 43. Ito Y, Bae SC, Chuang LS. The RUNX family: developmental regulators in cancer. Nat Rev Cancer 2015;15:81–95. 44. Kuleshov MV, Jones MR, Rouillard AD, Fernandez NF, Duan Q, Wang Z, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res 2016;44:W90–7. 45. Thannickal VJ, Fanburg BL. Activation of an H2O2-generating NADH oxidase in human lung fibroblasts by transforming growth factor beta 1. J Biol Chem 1995;270:30334–8. 46. Gutierrez S, Javed A, Tennant DK, van Rees M, Montecino M, Stein GS, et al. CCAAT/enhancer-binding proteins (C/EBP) beta and delta activate osteocalcin gene transcription and synergize with Runx2 at the C/EBP element to regulate bone-specific expression. J Biol Chem 2002; 277:1316–23. 47. Chuang LS, Ito K, Ito Y. RUNX family: Regulation and diversification of roles through interacting proteins. Int J Cancer 2013;132: 1260–71. 48. Wang CQ, Krishnan V, Tay LS, Chin DW, Koh CP, Chooi JY, et al. Disruption of Runx1 and Runx3 leads to bone marrow failure and leukemia predisposition due to transcriptional and DNA repair defects. Cell Rep 2014;8:767–82. 49. Krishnan V, Ito Y. A regulatory role for RUNX1, RUNX3 in the maintenance of genomic integrity. Adv Exp Med Biol 2017;962:491–510. 50. Glasauer A, Chandel NS. Targeting antioxidants for cancer therapy. Biochem Pharmacol 2014;92:90–101. Cancer Research Downloaded from http://aacrjournals.org/cancerres/article-pdf/78/1/88/2765428/88.pdf by guest on 26 June 2022 9. Kirshner J, Jobling MF, Pajares MJ, Ravani SA, Glick AB, Lavin MJ, et al. Inhibition of transforming growth factor-beta1 signaling attenuates ataxia telangiectasia mutated activity in response to genotoxic stress. Cancer Res 2006;66:10861–9. 10. Wang M, Saha J, Hada M, Anderson JA, Pluth JM, O'Neill P, et al. Novel Smad proteins localize to IR-induced double-strand breaks: interplay between TGFbeta and ATM pathways. Nucleic Acids Res 2013;41:933–42. 11. Comaills V, Kabeche L, Morris R, Buisson R, Yu M, Madden MW, et al. Genomic instability is induced by persistent proliferation of cells undergoing epithelial-to-mesenchymal transition. Cell Rep 2016; 17:2632–47. 12. Ito Y. RUNX genes in development and cancer: regulation of viral gene expression and the discovery of RUNX family genes. Adv Cancer Res 2008;99:33–76. 13. Yano T, Ito K, Fukamachi H, Chi XZ, Wee HJ, Inoue K, et al. The RUNX3 tumor suppressor upregulates Bim in gastric epithelial cells undergoing transforming growth factor beta-induced apoptosis. Mol Cell Biol 2006;26:4474–88. 14. Chi XZ, Yang JO, Lee KY, Ito K, Sakakura C, Li QL, et al. RUNX3 suppresses gastric epithelial cell growth by inducing p21(WAF1/Cip1) expression in cooperation with transforming growth factor {beta}-activated SMAD. Mol Cell Biol 2005;25:8097–107. 15. Whittle MC, Izeradjene K, Rani PG, Feng L, Carlson MA, DelGiorno KE, et al. RUNX3 controls a metastatic switch in pancreatic ductal adenocarcinoma. Cell 2015;161:1345–60. 16. Redon CE, Weyemi U, Parekh PR, Huang D, Burrell AS, Bonner WM. gamma-H2AX and other histone post-translational modifications in the clinic. Biochim Biophys Acta 2012;1819:743–56. 17. Harper JW, Elledge SJ. The DNA damage response: ten years after. Mol Cell 2007;28:739–45. 18. Lindahl T. Instability and decay of the primary structure of DNA. Nature 1993;362:709–15. 19. Ryoo IG, Ha H, Kwak MK. Inhibitory role of the KEAP1-NRF2 pathway in TGFbeta1-stimulated renal epithelial transition to fibroblastic cells: a modulatory effect on SMAD signaling. PLoS One 2014;9:e93265. 20. Jain M, Rivera S, Monclus EA, Synenki L, Zirk A, Eisenbart J, et al. Mitochondrial reactive oxygen species regulate transforming growth factor-beta signaling. J Biol Chem 2013;288:770–7. 21. Hubackova S, Kucerova A, Michlits G, Kyjacova L, Reinis M, Korolov O, et al. IFNgamma induces oxidative stress, DNA damage and tumor cell senescence via TGFbeta/SMAD signaling-dependent induction of Nox4 and suppression of ANT2. Oncogene 2016;35:1236–49. 22. Bakin AV, Stourman NV, Sekhar KR, Rinehart C, Yan X, Meredith MJ, et al. Smad3-ATF3 signaling mediates TGF-beta suppression of genes encoding Phase II detoxifying proteins. Free Radic Biol Med 2005; 38:375–87. 23. Liu RM, Choi J, Wu JH, Gaston Pravia KA, Lewis KM, Brand JD, et al. Oxidative modification of nuclear mitogen-activated protein kinase phosphatase 1 is involved in transforming growth factor beta1-induced expression of plasminogen activator inhibitor 1 in fibroblasts. J Biol Chem 2010;285:16239–47. 24. Gatza ML, Lucas JE, Barry WT, Kim JW, Wang Q, Crawford MD, et al. A pathway-based classification of human breast cancer. Proc Natl Acad Sci U S A 2010;107:6994–9. 25. Huang RY, Wong MK, Tan TZ, Kuay KT, Ng AH, Chung VY, et al. An EMT spectrum defines an anoikis-resistant and spheroidogenic intermediate mesenchymal state that is sensitive to e-cadherin restoration by a src-kinase inhibitor, saracatinib (AZD0530). Cell Death Dis 2013;4:e915. 26. Tan TZ, Miow QH, Huang RY, Wong MK, Ye J, Lau JA, et al. Functional genomics identifies five distinct molecular subtypes with clinical relevance and pathways for growth control in epithelial ovarian cancer. EMBO Mol Med 2013;5:983–98. 27. Voon DC, Wang H, Koo JK, Nguyen TA, Hor YT, Chu YS, et al. Runx3 protects gastric epithelial cells against epithelial-mesenchymal transition-induced cellular plasticity and tumorigenicity. Stem Cells 2012; 30:2088–99.