Interferon‐regulated suprabasin is essential for stress‐induced stem‐like cell conversion and therapy resistance of human malignancies

Radiation and chemotherapy represent standard‐of‐care cancer treatments. However, most patients eventually experience tumour recurrence, treatment failure and metastatic dissemination with fatal consequences. To elucidate the molecular mechanisms of resistance to radio‐ and chemotherapy, we exposed human cancer cell lines (HeLa, MCF‐7 and DU145) to clinically relevant doses of 5‐azacytidine or ionizing radiation and compared the transcript profiles of all surviving cell subpopulations, including low‐adherent stem‐like cells. Stress‐mobilized low‐adherent cell fractions differed from other survivors in terms of deregulation of hundreds of genes, including those involved in interferon response. Exposure of cancer cells to interferon‐gamma but not interferon‐beta resulted in the development of a heterogeneous, low‐adherent fraction comprising not only apoptotic/necrotic cells but also live cells exhibiting active Notch signalling and expressing stem‐cell markers. Chemical inhibition of mitogen‐activated protein kinase/ERK kinase (MEK) or siRNA‐mediated knockdown of extracellular signal‐regulated kinase 1/2 (Erk1/2) and interferon responsible factor 1 (IRF1) prevented mobilization of the surviving low‐adherent population, indicating that interferon‐gamma‐mediated loss of adhesion and anoikis resistance required an active Erk pathway interlinked with interferon signalling by transcription factor IRF1. Notably, a skin‐specific protein suprabasin (SBSN), a recently identified oncoprotein, was among the top scoring genes upregulated in surviving low‐adherent cancer cells induced by 5‐azacytidine or irradiation. SBSN expression required the activity of the MEK/Erk pathway, and siRNA‐mediated knockdown of SBSN suppressed the low‐adherent fraction in irradiated, interferon‐gamma‐ and 5‐azacytidine‐treated cells, respectively, implicating SBSN in genotoxic stress‐induced phenotypic plasticity and stress resistance. Importantly, SBSN expression was observed in human clinical specimens of colon and ovarian carcinomas, as well as in circulating tumour cells and metastases of the 4T1 mouse model. The association of SBSN expression with progressive stages of cancer development indicates its role in cancer evolution and therapy resistance.

Radiation and chemotherapy represent standard-of-care cancer treatments. However, most patients eventually experience tumour recurrence, treatment failure and metastatic dissemination with fatal consequences. To elucidate the molecular mechanisms of resistance to radio-and chemotherapy, we exposed human cancer cell lines (HeLa, MCF-7 and DU145) to clinically relevant doses of 5-azacytidine or ionizing radiation and compared the transcript profiles of all surviving cell subpopulations, including low-adherent stem-like cells. Stress-mobilized low-adherent cell fractions differed from other survivors in terms of deregulation of hundreds of genes, including those involved in interferon response. Exposure of cancer cells to interferon-gamma but not interferon-beta resulted in the development of a heterogeneous, low-adherent fraction comprising not only apoptotic/necrotic cells but also live cells exhibiting active Notch signalling and expressing stem-cell markers. Chemical inhibition of mitogen-activated protein kinase/ ERK kinase (MEK) or siRNA-mediated knockdown of extracellular signal-regulated kinase 1/2 (Erk1/2) and interferon responsible factor 1

Introduction
Over the last decades, genotoxic therapies including chemo-or radiotherapy have been the prevailing modalities used to treat cancer patients. Unfortunately, the initial effectiveness of radiotherapy and chemotherapy is commonly undermined by the primary and/or acquired resistance of tumour cells associated with the development of recurrent disease and progression towards a largely incurable metastatic stage that is responsible for over 90% of cancer-related mortality. The mechanisms underlying the occurrence of cancer resistance to radiotherapy and chemotherapy are complex and still poorly understood. Accumulating evidence indicates that the resistance of cancer cells to radiotherapy and chemotherapy encompasses several different but partly interlinked mechanisms comprising therapy-induced molecular programs together with the initial intra-tumour genetic and epigenetic heterogeneity (Burrell et al., 2013;Kreso et al., 2013;Swanton, 2012).
Notably, recent research has provided evidence that the cellular response to genotoxic stress can contribute to phenotypic reprogramming of cancer cells into more resistant phenotypes. Such process commonly involves induction of epithelial-to-mesenchymal transition (EMT) and occurrence of cancer cells with stem-like characteristics (Ghisolfi et al., 2012;Gomez-Casal et al., 2013;Kyjacova et al., 2015;Skvortsova et al., 2008). Generally, cancer-induced mesenchymal and stem-like cells appear to be more resistant to genotoxic stress compared with epithelial cancer cells (Blanpain et al., 2011). In addition, genotoxic therapies increase the migratory and invasive properties of malignant cells (Moncharmont et al., 2014) and alter the tumourassociated microenvironment, collectively promoting metastatic behaviour of treatment-surviving cancer cells (Ruegg et al., 2011).
There is mounting evidence that EMT is coupled with activation of the stem cell programme in both normal (Mani et al., 2008) and transformed cells (Han et al., 2013;Rhim et al., 2012). Whereas cancer stem cells (CSC) express EMT transcription factors such as Twist1, Snail and Slug, in contrast, EMT-undergoing cells were found enriched for CSC markers (Thiery et al., 2009). Related to the non-responsiveness of CSC to the currently used treatment modalities, some authors detected putative CSC following the therapy as the only residual cell population (Levina et al., 2008). Suprabasin (SBSN) was originally found in the suprabasal layer of murine and human keratinocytes and was described as an epidermal differentiation marker and component of keratinocyte cornified envelope and as a substrate for transglutaminase-mediated cross-linking (Alam et al., 2014;Park et al., 2002). However, its significant upregulation after increased demethylation of CpG islands in promoter of nonsmall cell lung carcinoma (Glazer et al., 2009) and salivary gland adenoid cystic carcinoma has been described , and later experiments revealed a role of SBSN in tumour cell proliferation and invasiveness (Zhu et al., 2016). These studies collectively suggest that SBSN is a new candidate oncogene contributing to tumour development and progression. Although Brother of the Regulator of Imprinted Sites factor has been found to contribute to regulation of SBSN expression , detailed mechanistic insights into regulation of SBSN expression in tumours is still missing.
Recently, we have shown that exposure of human metastasis-derived prostate cancer (PCa) cell lines to a clinically relevant fractioned ionizing radiation (fIR) regimen resulted in development of three distinct fIRsurviving populations in vitro : (1) senescent-like adherent cells with the potential to restart proliferation after the last dose of IR; (2) lowadherent dormant cells with stem-like cell traits and preserved competence to re-attach and restore proliferation (low-adherent cells); and (3) re-adherent cells derived from the low-adherent fraction with reverted epithelial features and retained tumourigenicity supporting a model of bidirectional interconversion in establishment of tumour cell heterogeneity (reviewed in van Neerven et al., 2016). We observed the involvement of an EMT-like programme in the response of PCa cells to fIR with such a phenotypic switch strongly dependent on the extracellular signal-regulated kinase 1/2 (Erk1/2)-mediated expression of Snail and associated with repression of E-cadherin. However, the basis for the heterogeneous response of cancer cells to genotoxic stress remains poorly understood.
In this study, we explored the signalling pathways involved in phenotypic remodelling of cancer cells induced by radiotherapy and/or chemotherapy. Comparing transcription and proteomic profiles of three human cancer models of interconverting subpopulations of stress-surviving tumour cells we identified a hierarchy of interferon and Erk signalling pathways in the regulation of SBSN and its novel function in establishment of the phenotypic plasticity and heterogeneity of therapy-resistant cancer cell populations.
DU145, MCF-7 and U373 cells were irradiated 24 h after seeding (22 000 cellsÁcm À2 ) with orthovoltage X-ray instrument T-200 (Wolf-Medizintechnik, St. Gangloff, Germany) using 0.5 GyÁmin À1 dose rate and thorium filter either daily with 10 doses 2 Gy or with a single dose of 2 or 10 Gy . Half volume of culture medium was changed every 48 h for a fresh one. All low-adherent cells were collected by centrifugation (300 g for 10 min) and returned back into cultivation flask. Twenty-four hours after the last dose of fIR and 72 h after a single dose, low-adherent and adherent cells were harvested separately.
To treat cells with 5-AC, HeLa, HS-5 and MCF-7 cells were seeded at a density of 22 000 cellsÁcm À2 24 h before 5-AC addition. A stock solution of 5-AC in sterile water was added either directly into culture medium (day 1, 3, 5, 7) or with exchange of a half volume of culture medium (day 2, 4 and 6) to a final concentration 2 or 4 lM with or without MEK inhibitor selumetinib (1 lM). Low-adherent cells were always re-collected as in the case of fIR. Low-adherent and adherent cells were harvested separately at day 8. Alternatively, a single dose of 5-AC was used and both low-adherent and adherent cells were harvested 72 h later (see Fig. 1A).
To treat cells with interferons, HeLa or MCF-7 cells were seeded at a density of 22 000 cellsÁcm À2 . Next day the culture medium was exchanged for a fresh one with or without IFNc (5 ngÁmL À1 ) or IFNb (50 ngÁmL À1 ) and cells were incubated for an additional 3 (HeLa) or 6 days (MCF-7). In the case of Erk pathway inhibition, MEK inhibitor selumetinib (1 lM) was added together with the first addition of IFNc and then every 24 h either directly to culture medium or with medium exchange (half volume) made every 48 h with re-collection of all low-adherent cells as described above.

Human colon carcinoma and ovarian cancer samples
See Table S3 for patient characteristics. Informed written consent has been obtained from all patients in compliance with the Declaration of Helsinki. The study was approved by the Ethics Committee of the Third Faculty of Medicine, Charles University, Prague, Ethical Commission of the National Institute of Public Health in Prague and by the Ethics committee of the General University Hospital Prague.

4T1 mouse metastatic model
Six-week-old BALB/c mice (females) were injected with 1 9 10 6 4T1 cells in PBS into mammary gland. Mice were sacrificed after 3 weeks when tumours on average reached AE 800 mm 3 ; lungs, liver and blood were removed and processed according to protocol described by Pulaski et al. (2001). To determine the level of SBSN expression, 4T1 colonies selected by 6-thioguanine were harvested and processed as described below in section on Quantitative real time PCR. Experiments were approved by the Czech Academy of Sciences Ethics Committee and performed according to the Czech Council guidelines for the Care and Use of Animals in Research and Teaching.

Colony-forming assay
To estimate the potential of low-adherent cells to reestablish adherence, the modified colony-forming assay was used as reported . (B) Loss of adhesion expressed as a ratio (fold induction) of Annexin V À /Hoechst À detached cells relative to all Annexin V À /Hoechst À cells (see Section 2 Materials and methods) assessed by FACS in control cells or cells treated with 2 lM 5-AC for 7 days. (C) RT-qPCR quantification of Sox2 and CD44 mRNA levels in parental, adherent and low-adherent HeLa cells treated with 4 lM 5-AC for 7 days. (D) RT-qPCR quantification of DLL1, DLL4, Hes1 and Hey1 mRNA levels in parental, adherent and low-adherent HeLa cells treated with 4 lM 5-AC for 7 days. (E) Immunoblotting detection of Snail in parental, adherent and low-adherent HeLa and MCF-7 cells treated with 4 lM 5-AC for 72 h. GAPDH was used as a loading control. (F) RT-qPCR quantification of p16 and p21 mRNA levels in parental, adherent and low-adherent HeLa cells treated with 2 lM 5-AC for 72 h. (G) Immunoblotting detection of Bcl-XL in parental, adherent and lowadherent HeLa cells treated with 4 lM 5-AC for 72 h. Gamma-tubulin was used as a loading control. (H) Estimation of low-adherent fraction after inhibition of Erk kinase pathway with selumetinib (ERKi, 1 lM) in combination with siRNA knockdown of Bim (siBim) expressed as a ratio (in fold induction) of Annexin V À /Hoechst À detached cells relative to all Annexin V À /Hoechst À cells assessed by FACS in HeLa cells treated with 2 lM 5-AC for 72 h. Non-targeting siRNA (siNC) was used as a control. (I) Immunoblotting detection of Erk phosphorylated on threonine 202/tyrosine 204 and Bim in HeLa cells treated with 2 lM 5-AC for 72 h after knockdown of Bim (siBim) or in the presence of MEK/Erk inhibitor selumetinib (ERKi; 1 lM). GAPDH was used as a loading control. Data are shown as mean values AE SEM, with n = 3 per group. Two-way ANOVA was used for multigroup comparisons followed by Tukey's post hoc test. The asterisk represents P < 0.05. Briefly, 3 days after the exposure to treatment, low-adherent cells were transferred onto a new plate and left for 24 h to allow adherence of residual mitotic cells. The remaining low-adherent cells were transferred to a new plate and allowed to re-adhere for 24-30 days. To visualize formation of colonies, cells were fixed at À20°C with 96% ice-cold methanol for 10 min and then stained with 0.5% crystal violet in ethanol at room temperature for 5 min. 2.6. Indirect immunofluorescence U373 cells were transfected by siRNA 24 h after seeding at a density of 22 000 cellsÁcm À2 on glass coverslips. Culture medium was changed 24 h after transfection. Cells were irradiated by a single dose of 10 Gy 48 h post-transfection. Seventy-two hours postirradiation, cells were fixed with 4% formaldehyde for 15 min and then permeabilized by 0.1% Triton X-100. After washing with PBS, cells were incubated in 10% FBS/1 9 PBS for 60 min, with primary antibody at room temperature for 1 h, and after PBS wash with secondary antibodies at room temperature for 1 h. After washing with 1 9 PBS and ddH 2 O, cells were stained with 1 lgÁmL À1 DAPI (Sigma), washed with ddH 2 O and mounted with ProLong TM Gold Antifade Mountant (Thermo Fisher Scientific). Images were acquired by a Leica DM6000 fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
Low-adherent MCF-7 cells were generated using a single dose of 4 lM 5-AC. After 5 weeks with medium exchange every 3-4 days, cells were washed with PBS and cytospin onto microscopic slides (400 g for 5 min; Centurion Scientific, K3 Series, Irvine, CA, USA). Attached cells were fixed using 4% formaldehyde for 15 min and then permeabilized by 0.1% Triton X-100. After washing with PBS and ddH 2 O, cells were stained with 1 lgÁmL À1 DAPI (Sigma), washed with ddH 2 O and mounted with ProLong TM Gold Antifade Mountant (Thermo Fisher Scientific). Images were acquired with a Leica TCS SP8 confocal laser microscopy system (Leica Microsystems).

Magnetic-activated cell sorting (MACS)
Cells were treated for 72 h and low-adherent cells were collected 24 h after the last dose of the particular treatment. The Annexin V-negative fraction of low-adherent human cancer cells was obtained by incubation of low-adherent cells with Dead Cell Removal MicroBeads (Dead Cell Removal Kit, Miltenyi Biotec, Bergisch Gladbach, Germany) for 15 min and separated using the magnetic field of an AutoMACS Pro magnetic separator (Miltenyi Biotec). The separated cells were processed immediately for immunoblotting, qRT-PCR or whole genome expression analysis.

Detection of apoptosis using fluorescenceactivated cell sorting (FACS)
The numbers of viable and dead cells were analyzed using an LSRII flow cytometer (BD Biosciences). Cells (adherent in one vial and dead plus low-adherent in other vial) were collected (400 g at 4°C for 5 min), washed with PBS and stained for 15 min at room temperature with 0.3 lL of Annexin V-Dyomics 647 in 100 lL of 1 9 Annexin V binding buffer (Apronex, Prague, Czech Republic). Immediately before analysis, 0.3 lL of Hoechst 33258 (10 mM; Invitrogen, Carlsbad, CA, USA) was added. The percentage of lowadherent cells (Annexin V À /Hoechst À ) was calculated according to the formula low-adherent % = [(low-adherent cells/(low-adherent cells + live-adherent cells)]*100. The calculated percentages of surviving low-adherent cells after the treatments were further normalized to the percentage of low-adherent cells in control conditions.

SDS/PAGE and immunoblotting
Cells were washed twice with PBS, harvested into Laemmli SDS sample lysis buffer (2% SDS, 50 mM Tris-Cl, 10% glycerol in double-distilled H 2 O) and sonicated (3 9 15 s at 4 lm amplitude with 15-s cooling intervals) using Soniprep 150 (MSE, London, UK). Conditioned culture media were precipitated by acetone overnight and samples then centrifuged (4000 g, 10 min), and air-dried pellets were dissolved into Laemmli SDS sample lysis buffer and sonicated (3 9 15 s at 4 lm amplitude with 15-s cooling intervals). Concentration of proteins was estimated by the BCA method (Pierce Biotechnology, Rockford, IL, USA). DTT 100 mM and 0.01% bromophenol blue were added to cell lysates before separation by SDS/ PAGE. The same amount of protein (35-50 lg) was loaded into each well. Proteins were electrotransferred onto a nitrocellulose membrane using wet transfer and detected by specific antibodies combined with horseradish peroxidase-conjugated secondary antibodies (goat anti-rabbit, goat anti-mouse; Bio-Rad). Peroxidase activity was detected by ECL (Pierce Biotechnology) or SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific). GAPDH, c-tubulin or b-actin was used as a marker of equal loading.

Quantitative real time PCR (RT-qPCR)
Total RNA samples were isolated using RNeasy Mini Kit (Qiagen Sciences, Germantown, MD, USA) as described (Vlasakova et al., 2007). First strand cDNA was synthesized from 500 ng of total RNA with random hexamer primers using a High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Foster City, CA, USA). RT-qPCR was performed in ABI Prism 7300 (Applied Biosystems) using SYBR Select Master Mix containing SYBR GreenE dye (Applied Biosystems). The relative quantity of cDNA was estimated by the DDCT method and data were normalized to RPL37a or b-actin (ACTB).
The following primers were purchased from Eastport ( The data are expressed as the means AE SEM of a minimum of three independent experiments performed in triplicate. The P-values were estimated using twotailed Student's t-test. P-values < 0.05 were considered statistically significant.

Transient transfection and dual-luciferase reporter assay
HEK293 were seeded at a density of 22 000 cellsÁcm À2 onto a 60-mm dish. Twenty-four hours later the medium was exchanged and the cells were co-transfected with 6 lg pGL3-SBSN, pGL-3-Control (SV40 promoter) or pGL3-Basic (promoterless) vector encoding Firefly luciferase together with 6 lg pRL-TK vector encoding Renilla luciferase using the calcium phosphate method. After medium exchange 24 h post-transfection, cells were treated with IFNc (5 ngÁmL À1 ) or PMA (20 nM). MEK inhibitor selumetinib (1 lM) was added to cell cultures either alone or simultaneously with addition of IFNc. Samples were collected 48 h after treatments into 1 9 Passive lysis buffer and further processed according to manufacturer's protocol (Promega). Luciferase activity was detected by 2104 EnVision Multimode plate reader (Perkin Elmer, Waltham, Massachusetts, USA). Luminometric data of Firefly luciferase activity were normalized to Renilla luciferase activity and protein concentration determined by BCA assay (Thermo Fisher Scientific).

Whole-genome expression array
To obtain fIR-resistant cell fractions (10 9 2 Gy daily), DU145 and MCF-7 cells were seeded at a density of 14 000 cellsÁcm À2 into two 175-cm 2 cultivation flasks 24 h before first dose. Low-adherent cells were recollected every medium exchange (48 h). Low-adherent cells were separated from adherent ones 24 h after the last dose. One half of low-adherent cells was harvested (low-adherent fraction; see below) in parallel with adherent cells (adherent fraction) and the remaining low-adherent cells were kept in culture to form re-adherent colonies and harvested after 21 days (readherent fraction). Proliferating cells served as controls (parental cells).
To generate 5-AC-resistant cell fractions (seven times the daily dose of 5-AC), HeLa cells were seeded at a density of 14 000 cellsÁcm À2 into two 175-cm 2 cultivation flasks 24 h before first addition of 5-AC (4 lM) and cultured as described above. Twenty-four hours after the last dose of 5-AC, half of low-adherent cells was harvested in parallel with adherent cells. Re-adherent cells were prepared in a similar manner as those for fIR. Proliferating cells served as control (parental fraction).
Annexin V-negative (live) low-adherent cells were separated from dead cells by MACS as described above. Total RNA of all cell fractions (parental, lowadherent, adherent and re-adherent) was isolated using RNeasy Mini Kit (Qiagen Sciences). RNA integrity was assessed in Agilent 2100 Bioanalyzer and RNA 6000 Nano LabChip (Agilent Technologies, Santa Clara, CA, USA). Total RNA was amplified using Illumina TotalPrep RNA Amplification Kit from Applied Biosystems according to the standard protocol, from a starting amount of 200 ng. The cDNA quality and quantity were assessed in Agilent 2100 Bioanalyzer and RNA 6000 Nano LabChip. The cDNA (750 ng) was hybridized, washed and scanned according to the manufacturer's instructions.
The raw expression data resulting from microarray were analyzed using the BeadArray package (Dunning et al., 2007) of Bioconductor within the R environment (R Development Core Team 2007). All hybridizations passed the quality control. The data were backgroundcorrected and normalized with the probe level quantile method. Differential expression analysis was performed with the Limma package (Smyth et al., 2005) on intensities that were variance-stabilized by logarithmic transformation. Annotation provided by Bioconductor was used (illuminaHumanv4BeadID.db). To identify significantly perturbed pathways, we performed SPIA (Tarca et al., 2009) analysis on KEGG pathways: genes with Pvalues < 0.05 were considered to be differentially transcribed. RNA-seq data have been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/ arrayexpress) under accession number E-MTAB-6062.

Proteomic analysis
DU145 cells for proteomic analysis were cultivated in 'heavy stable isotope labelling with amino acids in cell culture (SILAC)' low glucose medium (Sigma, D9443) supplemented with L-arginine-13C6,15N4 (84 mgÁmL À1 ), L-lysine 13C6 (146 mgÁmL À1 ), L-leucine (0.1 mgÁmL À1 ), sodium pyruvate (1 mgÁmL À1 ), D-glucose (25 mM) and 10% dialyzed FBS (Sigma, F0392). Ninety-five percent efficiency of labelling was achieved after 10 population doublings in heavy media. Low-adherent cells were prepared according to previously published protocol . Cells grown on T150 tissue culture flasks were irradiated with orthovoltage X-ray instrument T-200 (Wolf-Medizintechnik GmbH) using 0.5 GyÁmin À1 dose rate and thorium filter for 10 days with 2 Gy (fIR). Every third day, medium was collected and centrifuged. Pelleted cells were resuspended in fresh medium and returned to the original dish. After the last dose, the medium containing low-adherent cells was transferred to a fresh flask and incubated for additional 24 h to allow the mitotic cells to re-adhere. Medium was centrifuged at 300 g for 10 min to pellet the low-adherent cells. Annexin Vnegative fraction was obtained by incubation with Dead Cell Removal MicroBeads (Dead Cell Removal Kit; Miltenyi Biotec) for 15 min and separation in magnetic field of AutoMACS Pro magnetic separator (Miltenyi Biotec).
Cells from three separations were lysed in lysis buffer (100 mM 4-ethylmorpholine pH 8.3, 10% acetonitrile, 1% w/v sodium deoxycholate). Lysates were heated to 95°C for 5 min and sonicated until clear, non-viscous solution was obtained. Total protein concentration was measured by BCA assay (Thermo, cat No 23225). Lysates from heavy labelled cells (floating population) and nonlabelled cells (adherent non-irradiated cells) were mixed in a 1 : 1 ratio to obtain a total of 60 lg of protein. Samples were reduced with neutralized 5 mM TCEP (Thermo, 77720) at 60°C for 15 min followed by alkylation of cysteines with 10 mM MMTS, at room temperature for 30 min. Proteins were digested with trypsin GOLD (Promega, V5280) at 1 : 40 ratio (enzyme : protein) at 37°C for 16 h.
Digested proteins were acidified by addition of 4 lL of 40% TFA. Precipitated deoxycholate was extracted five times with ethyl acetate. Insoluble particles were spun at 16 000 g for 15 min and the supernatant was dried using a speed vac.
Raw data were processed using MAXQUANT (Cox and Mann, 2008). Statistical analysis and annotation enrichment analyses were performed in PERSEUS SOFTWARE (Tyanova et al., 2016).

Data processing and statistical analysis
Unless stated otherwise, data are mean values AE standard error of means (SEM) of at least three independent experiments. In mouse experiments, groups of six animals were used, unless stated otherwise. Two-way ANOVA presented as mean AE SEM was used to assess statistical significance with P < 0.05 being regarded as significant, using GRAPHPAD PRISM software, San Diego, CA, USA. Images are representative of at least three independent experiments. Fluorescence-activated cell sorting (FACS) data were analyzed using FLOWJO 9.6.4 cytometric analytical software (FlowJo LLC., Ashland, OR, USA).

5-Azacytidine induces loss of adhesion and stem cell traits associated with resistance to anoikis
The exposure of human cancer cells to fIR resulted in development of resistant cell subpopulations including cells losing adhesion and bearing CSC-like features . To find out whether the development of such low-adherent cell fraction represents a common response of tumour cells to genotoxic stress, we exposed human cancer cell lines HeLa and MCF-7, and retrovirally immortalized bone marrow mesenchymal stem cells HS-5 to 5-azacytidine (5-AC), a hypomethylating agent currently used in the therapy of high-risk myelodysplastic syndrome (MDS; Kornblith et al., 2002) and tested in numerous clinical trials for treatment of solid malignancies. We used two modes of 5-AC exposure (as specified for each experiment in the figure legends): (1) a 'clinical-mimicking' regimen of 5-AC with daily administration for seven consecutive days with cells harvested at day 8 (see Fig. 1A, left), and (2) a single dose of 5-AC with harvest at day 3 showing a similar effect on resistance development achieved in shorter time (see Fig. 1A, right), both within the concentration range of 5-AC that is reached in the humans (in MDS patients 5-AC is administered at a dose of 75 mgÁm À2 Áday À1 ; (Marcucci et al., 2005;Rudek et al., 2005). Both 5-AC-treatments mostly reproduced the effects of fIR published previously , showing the equivalence of both 5-AC treatments. Besides senescence-like state (not shown) and cell death (see Fig. S1A), 5-AC treatment induced development of a low-adherent cell subpopulation (see Fig. S1B), the onset of which was detectable in all three cell types from day 2 onwards (see Fig. 1B). Comparable to fIR, 5-AC-induced occurrence of anoikis-resistant cells that survived in a low-adherent (dormant) state for 2-4 weeks before readhesion and renewal of proliferation ( Fig. S1C; only MCF-7 cells shown). The lowadherent cells exhibited stem cell-like traits such as elevation of SRY (sex determining region Y)-box 2 (Sox2), CD44 and OCT4 transcripts (see Fig. 1C for HeLa cells and Fig. S1D,E for MCF-7 and HS-5 cells), activation of Notch signalling (detected as elevation of Notch ligands DLL1 and DLL4 and Notch effector transcription factors Hes1 and Hey1; see Fig. 1D for HeLa cells and Fig. S1F,G for MCF-7 and HS-5 cells) and elevation of Snail (see Figs 1E and S1H for protein and mRNA levels of Snail, respectively). Furthermore, the lack of proliferation in the low-adherent cell fraction correlated with increased mRNA levels of inhibitors of cyclin-dependent kinases p16 INK4a (p16) and p21 waf1 (p21; Fig. 1F for HeLa cells and Fig. S1I,J for MCF-7 and HS-5 cells; note, the mRNA levels of cell cycle cyclin-dependent kinases were significantly downregulated as detected by transcriptome analysis). In addition, the level of anti-apoptotic factor Bcl-XL was most pronounced in the low-adherent fraction (Fig. 1G). Together with suppression of protein levels of E-cadherin and integrin aV (Fig. S1K) this demonstrated that 5-AC is capable of inducing phenotypic changes reminiscent of those evoked by fIR . The activation of kinases Erk1/2 (detected as phosphothreonine 202 and phosphotyrosine 204 of Erk1 and Erk2, respectively) and Akt (detected as phosphoserine 473) was observed (Fig. S1L). Consistent with a previous study , inhibition of Erk1/2 activity by MEK inhibitor selumetinib suppressed survival of the low-adherent fraction and reconstitution of fully adherent colonies (Fig. S1M). Of note, siRNA-mediated knockdown of Bim, a protein involved in anoikis, whose degradation is controlled by Erk activity, resulted in preservation of viability in low-adherent cells even in the presence of selumetinib ( Fig. 1H; for efficiency of Bim downregulation, see Fig. 1I), supporting the role of the Erk/Bim pathway in the survival of low-adherent cells.
Altogether, these data demonstrate that 5-AC induces phenotypic plasticity characterized by survival of low-adherent tumour cells with stem cell-like features in different human cell types in a manner similar to fIR, thus indicating that this type of cell response is shared by various genotoxic insults.

Elevated interferon response is a common feature of surviving low-adherent cells
To characterize further the low-adherent cells surviving 5-AC and radiation treatments, a comparative analysis of gene expression profiles of irradiated cells (DU145 and MCF-7; 10 9 2 Gy) with 5-AC-treated cells (HeLa; 7 9 daily dose of 4 lM 5-AC) was performed using Illumina human gene expression array. The hierarchical clustering of resistant populations revealed 2449 commonly deregulated protein-coding genes in low-adherent survivors ( Fig. 2A, Table S1A). The heat map representation of the clustered matrix showed enrichment of genes involved in the innate immune response pathways (cluster I), specifically, nuclear factor jB (NF-jB), Toll-like receptor, and interferon signalling (for details, see Figs 2B and S2A, Table S1B). The changes of mRNA levels of several interferon stimulated genes (IRF1, IRF7, ISG15, Her-c5, SOCS-1 and MX-1) were validated by RT-qPCR in HeLa cells treated with 5-AC for 7 days (Fig. 2C). Cluster II predominantly contained genes involved in the mitochondrial biogenesis and maintenance. Finally, cluster III contained genes involved in the regulation of transcription. These genes were, conversely, upregulated in all stress-induced low-adherent cells in comparison with the non-treated controls (Table S1B).
To obtain a more comprehensive view of the biological state of low-adherent cells, we performed proteomic profiling of irradiated (10 9 2 Gy) DU145 cells using SILAC and mass spectrometry (Ong et al., 2002; Table S2A). As a next step, the protein expression data were correlated with mRNA expression profiles of fIR low-adherent DU145 cells using 2D annotation enrichment (Cox and Mann, 2012;Fig. S2B, Table S2B). Analogous to gene expression profiles, the proteomic pattern in the fIR low-adherent DU145 cells was dominated by the 'interferon    response' pathway including proteins encoded by interferon stimulated genes such as HLA-C, OAS1 and OAS3, ISG15 and ISG20, MX1 and STAT1 (Der et al., 1998;Fig. S2C; for relative expressions of genes containing this annotation see Fig. 2D).
To conclude, both fIR-and 5-AC-treatment induced development of the low-adherent surviving cell fraction characterized by profound changes in the expression of genes and proteins involved mainly in interferon response and cell cycle regulation.

IFNc but not IFNb induces loss of adhesion and anoikis-resistant survival of human cancer cells
Based on the transcriptome and proteomic profiling, we next tested whether the activity of the interferon pathway is causally involved in the development of low-adherent state and survival of cancer cells. For this purpose, HeLa and MCF-7 cells were exposed to daily administration of interferon type I (IFNb) and type II (IFNc) for 3 or 6 days, and the development of the low-adherent cell subpopulation was followed.
In contrast to IFNb, IFNc significantly induced lowadherent population from day 3 in HeLa and day 6 in MCF-7 cells (Fig. 3A). Notably, the administration of IFNc induced the loss of adhesion in HS-5 cells as well (Fig. S3A). As expected, besides STAT1 receptor-mediated phosphorylation (detected as STAT1 tyrosine 701 phosphorylation), the response to IFNb and IFNc was reflected by induction of transcription factors involved in the late interferon response IRF7 and IRF1, respectively (see Fig. 3B for HeLa and 3c for MCF-7 cells). Formation of large extracellular vesicles (see Fig. S3B; Rak, 2010) in both HeLa and MCF-7 cells was associated with a switch to the amoeboid-like form of cell migration (Di Vizio et al., 2009) of lowadherent cells (as shown in Video S1). Interestingly, some IFNc-induced low-adherent cells were able to re-adhere within several days after the end of the treatment and to reestablish adherent cell growth (Fig. 3H, left). Similarly to fIR, the extended exposure of cells to IFNc resulted in marked elevation of Sox2 and Oct4 mRNA levels in the low-adherent cell fraction ( Fig. 3D; HeLa cells shown) and activation of Erk1/2 (see Fig. 3E for HeLa and Fig. 3F for MCF-7 cells). Intriguingly, MEK inhibitor selumetinib (Fig. S3C) suppressed the IFNc-induced development of low-adherent cells (Fig. 3G) as well as the establishment of re-adherence (see Fig. 3H, right), indicating crosstalk between IFNc and MEK/ Erk signalling pathways in the anoikis-resistant cancer cell survival. Downregulation of late interferon response transcription factor IRF1 by RNA interference (Fig. S3D,E) resulted in suppression of the IFNc-induced (Fig. 3I) as well as IR-induced (Fig. 3J) low-adherent cell fraction, supporting the causal role of IFNc signalling in low-adherent survival after genotoxic stress. These data show that prolonged activation of the IFNc/JAK signalling pathway in human cancer cell models grown in vitro induces changes in cell adhesive properties and development of the low-adherent cell subpopulation matching phenotypic reprogramming observed after fIR and 5-AC administration, respectively. Altogether, these findings indicate that the development of radioresistant cell subpopulations is mediated by the concerted action of the IFNc/JAK and Erk signalling pathways.

Upregulation of suprabasin is a component of the shared response to stress induced by 5-azacytidine, fIR and IFNc
Furthermore, comparative analysis of transcript profiles of low-adherent cells induced by fIR (DU145 and MCF-7) and 5-AC (HeLa) revealed 463 genes whose transcripts were altered in the same direction (see Venn diagram in Fig. S4A). Notably, among the top scoring genes upregulated in surviving low-adherent cell fractions was SBSN (Fig. S4B), an oncoprotein associated with tumour progression and metastasis (Glazer et al., 2009;Shao et al., 2012;Zhu et al., 2016). Indeed, the elevation of SBSN transcript levels was confirmed in U373 cells on the mobilization of the low-adherent cell subpopulation after 2 Gy irradiation. The amount of low-adherent cells is expressed as a ratio (fold induction) of Annexin V À /Hoechst À detached cells relative to all Annexin V À /Hoechst À cells. Non-targeting siRNA (siNC) was used as a control. Data are shown as mean values AE SEM, with n = 3 per group. Two-way ANOVA was used for multigroup comparisons followed by Tukey's post hoc test. The asterisk represents P < 0.05.
To examine further the functional role of SBSN in radioresistance, the effect of SBSN downregulation on radiosensitivity of the U373 cell line was estimated. Indeed, knockdown of SBSN suppressed the development of IR-induced (2 Gy) low-adherent cell fraction (Fig. 4I) and increased the percentage of apoptotic cells (Fig. 4I), supporting the functional role of SBSN in radioresistance of glioblastoma cells. Altogether, our findings show that the expression of SBSN is a shared feature of the cell response to genotoxic stress induced by different stimuli and indicate that SBSN is required for the stress-induced phenotypic plasticity, cellular heterogeneity and resistance to apoptosis of tumour cell populations.
3.5. IFNc-and 5-AC-induced expression of SBSN is mediated by the MEK/Erk pathway Next we examined how the SBSN gene is regulated by IFNc and other stresses. As shown in Fig. 5A, RNA interference-mediated knockdown of IRF1 resulted in suppression of IFNc-induced elevation of the SBSN transcript in HeLa cells. As IFNc activated Erk1/2 (Fig. 3E,F) and MEK inhibitors affected IFNcinduced development of low-adherent fraction (Fig. 3G), we examined whether SBSN is regulated by IFNc-induced activation of Erk signalling. Downregulation of IRF1 suppressed IFNc-induced activity of Erk1/2 (Fig. 5B) and treatment of IFNc-exposed HeLa cells with selumetinib resulted in suppression of the IFNc-induced upregulation of SBSN (Fig. 5C) to a degree comparable to the impact of Erk1/2 downregulation ( Fig. 5E; see Fig. S5A for efficiency of Erk1/2 downregulation), indicating that Erk activity contributes to the observed upregulation of SBSN gene expression mediated by IFNc. The effect of inhibition of the Erk pathway (Fig. S5B) on the mRNA expression of SBSN was also found in HeLa and MCF-7 cells exposed to both 5-AC and selumetinib (Figs 5D and S5C,D; note MEK inhibitor U0126 was used to further support the data). Additionally, the downregulation of Erk1 and/or Erk2 also suppressed 5-ACinduced elevation of SBSN mRNA (see Fig. 5F; for Erk1/2 downregulation efficiency, see Fig. S5E), suggesting again the role of the Erk pathway in stressinduced SBSN expression.
Furthermore, a luciferase activity driven by the proximal regulatory region of the SBSN gene spanning nucleotides À2000 to +71 relative to the SBSN transcription start site (pGL3-SBSN) was estimated in HEK293 cells treated with either IFNc or MEK inhibitor, or both (see Section 2, Materials and methods). Indeed, IFNc treatment significantly induced Firefly luciferase expression driven by the SBSN proximal promoter relative to a promoterless construct (pGL3-Basic), whereas the MEK inhibitor selumetinib almost completely abolished it (Fig. 5G), suggesting the effect of IFNc on SBSN gene expression is mediated by Erk1/2 signalling. Moreover, an activator of MAPK signalling PMA induced SBSN promoter-driven Firefly luciferase expression, further supporting transcriptional control of SBSN by Erk1/2 in accordance with a previous study (Park et al., 2002).
Collectively, these findings indicate the crucial role of Erk kinase signalling in the upregulation of SBSN expression in cells exposed to genotoxic stress.

SBSN is expressed in malignant human tissues and is associated with metastasis
Recent reports indicate SBSN expression in human malignancies such as glioblastoma (Formolo et al., 2011) or lung cancer (Glazer et al., 2009). Indeed, the analysis of SBSN expression in colorectal cancer biopsies (n = 8) provided similar results (Fig. 6A). While normal mucosa surrounding the tumour tissue was always negative for SBSN mRNA expression, all tumour samples were positive (P < 0.05). Similarly, the SBSN mRNA levels in ovarian carcinoma biopsies ( Fig. 6B; n = 100) were significantly higher than in nonmalignant ovary tissue (n = 14, P < 0.0001).
Due to association of SBSN expression with EMT programme and cell stemness observed in our in vitro experiments, we next examined whether the SBSN expression is associated with development of metastases in vivo. Using a mouse model of 6-thioguanineresistant 4T1 cells that frequently metastasize to lungs and liver, we compared SBSN transcript levels in parental cells, primary tumours, 4T1-derived circulating tumour cells (CTCs), and lung and liver metastases (see Section 2 Materials and methods for details). Notably, the level of SBSN expression was significantly higher in CTC and lung and liver metastases than in primary tumour and parental 4T1 cells (Fig. 6C), a feature that correlated with increased Sox2 expression in these cells (Fig. 6D).
Altogether, SBSN is expressed in fully developed human malignant lesions of colorectal and ovarian carcinomas. Furthermore, the high expression of SBSN in circulating tumour cells and metastatic lesions detected in our mouse model suggests association of SBSN with the metastatic process.

Discussion
Metastatic dissemination frequently combined with development of resistance to current antitumour therapies represents the most serious complications of malignant diseases and causes more than 90% of deaths among cancer patients. The aim of this study was to elucidate the mechanisms of acquired resistance to radiotherapy and chemotherapy in several human cancer cell models, with emphasis on better understanding the development of a subpopulation of cancer cells characterized by anoikis-resistant survival. Among many deregulated genes detected by cDNA microarray analysis of transcript profiles from anoikisresistant human tumour cells induced by 5-azacytidine or fIR, we identified elevation of gene transcripts regulated by the interferon signalling pathway and IFNc itself (but not IFNb) as being capable of inducing the anoikis-resistant state. Furthermore, we pinpointed SBSN, the recently reported oncoprotein, as an effector factor involved in the development of cancer cell resistance phenotypes. Elevated expression of SBSN in mouse circulating tumour cells and metastases, as well as clinical samples of human colorectal and ovarian carcinomas, supported our in vitro findings that SBSN was involved in cancer development and progression.
Interferons are known to induce genotoxic stress (see Hubackova et al., 2015 and references therein). Here we observed that the exposure of tumour cells to IFNc but not IFNb was sufficient to mobilize a heterogeneous low-adherent and floating matter composed mainly of apoptotic and necrotic cell bodies, and a small fraction of cells that survived the IFNcinduced stress and later re-established adherent cell T1 T2 HM T1 T2 HM T1 T2 HM T1 T2 HM T1 T2  growth. These results indicate that interferon response itself or responses triggered by genotoxic insults alter cell adhesion properties in a subpopulation of cancer cells that is also associated with resistance to anoikis resulting in increased invasiveness of these cells.
As was shown previously  low-adherent cells with stem cell-like features are continuously generated from the adherent, senescent phenotype-developing fraction of resistant cells. Although the low-adherent cells partially share their gene expression pattern with senescent cells, for example elevated p21 waf1 and p16 INK4a , and several SASP factors, our comparison of global transcription profiles of lowadherent versus adherent cells indicates they are not identical. The major functional difference between the low-adherent and adherent cell fraction is evident from the ability of the former cells to re-attach and re-enter the cell cycle after a few weeks. This correlates with the recently published findings that a reprogrammed gene expression profile associated with de-differentiation, stem cell-like features, EMT, increased levels of c-myc and STAT, which we observed also in our lowadherent cells, is responsible for senescence bypass and therapy resistance (Deschênes-Simard et al., 2019). Furthermore, given that cytoplasmic DNA elicits IFN responses (primarily type I IFNb) and NF-jB signalling through the cGAS/STING pathway , we also searched the transcriptomes of the anoikis-resistant cell subpopulation for evidence of IFNb and NF-jB activity. Indeed, our data show increased mRNA levels of IFNB1, IL1, tumour necrosis factor (TNF), IL6 and IL18, common targets of the cGAS/ STING pathway. Future studies should address the possibility that the cGAS/STING pathway may contribute to development of the low-adherent phenotype and thereby play a role in survival and radiotherapyresistance of cancer cells.
Our analysis of IFNc-induced elevation of SBSN revealed that activation of Erk1/2 is a prerequisite for SBSN expression reminiscent of endothelial cells treated with epidermal growth factor as an activator of Erk pathway (Alam et al., 2014). Treatment of cells with IFNc resulted in increased Erk1/2 activity, the inhibition of which by MEK inhibitors reduced both SBSN expression and occurrence of low-adherent population after fIR, 5-AC or IFNc. Although we did not explore the mechanism of Erk activation by IFNc in depth, several reports show the ability of IFNc to activate the Erk pathway (see Cho et al., 2007;Halfter et al., 2005;Harris et al., 2007;Sakatsume et al., 1998;Stancato et al., 1998). Upon IFNc stimulation of keratinocytes, SOCS1 induced by JAK/STAT pathway inhibited IFNc receptor and STAT1 phosphorylation but maintained Erk activation (Madonna et al., 2008). Together with our observation of increased expression of SOCS1 in the low-adherent cells (Fig. 2C), these results implicate SOCS1 in modulation of therapy resistance via stabilization of Erk signalling.
Although the exact molecular function of SBSN remains currently unknown, there is emerging evidence that aberrant expression of SBSN is associated with human malignancies. SBSN has been found among genes coordinately expressed in non-small cell lung carcinoma characterized by promoter hypomethylation (Glazer et al., 2009). Intriguingly, ectopic expression of SBSN in some lung squamous cell carcinoma cell lines and normal lung fibroblasts is associated with growth advantage (Glazer et al., 2009). Likewise, SBSN promotes proliferation of oesophageal squamous cell carcinoma (ESCC) cell lines in vitro (Zhu et al., 2016). In addition, SBSN expression correlated with ESCC progression (i.e. SBSN expression correlated positively with increasing clinical tumour staging and negatively with overall patient survival time; Zhu et al., 2016). Furthermore, SBSN was found as part of the secretome of highly invasive glioblastoma cell lines (Formolo et al., 2011). Notably, aberrant expression of SBSN was found in pleural effusions of lung adenocarcinoma patients (Sheng and Zhu, 2014) and in tumour endothelial cells of renal and colon carcinomas (Alam et al., 2014). These recent findings of cancerassociated expression of SBSN were further supported and extended in our present study by detection of SBSN in colorectal and ovarian carcinomas.
The presence of SBSN transcript in both adherent and anoikis-resistant low-adherent cell populations, yet present predominantly in the latter cell fraction, suggested a potentially novel role of SBSN in the development of low-adherent state and anoikis-resistant survival. This scenario was indeed supported by the experiments manipulating SBSN levels, where the expression of SBSN dictated the amount of cells in the low-adherent state. SBSN overexpression activates the Wnt/b-catenin signalling pathway (Zhu et al., 2016), which has been implicated as a niche factor to maintain stem cells in a self-renewing state (Nusse, 2008). Wnt/b-catenin signalling is often associated with EMT as an activator of EMT transcription factors Zeb, Snail or Slug (Medici et al., 2008;Sanchez-Tillo et al., 2011). Thus, expression of SBSN in lowadherent cells can be responsible both for EMT due to increased expression of Snail and Zeb2 (Figs 1E and S1G), and for induction of cancer stemness (Figs 1C and S1C,D).
Based on these findings we propose the role of SBSN in stress resistance of cancer cells. It should be emphasized that SBSN may be involved in the maintenance of tumour invasiveness and metastatic potential, as was reported for salivary gland adenoid cystic carcinoma . The association of SBSN expression with tumour aggressiveness is underscored by our data indicating that SBSN induction is highest in circulating mouse tumour cells and metastatic deposits. Collectively, these findings implicate SBSN in promoting the aggressive phenotype of tumour cells.
Notably, Aibara et al. (2018) identified anti-SBSN antigen activity in immune complexes present in cerebrospinal fluid of patients with the neuropsychiatric form of systemic lupus erythematosus. This example of SBSN immunogenicity, documented by autoantibodies reactive with astrocytes causing an inflammatory response, may have implications for future attempts to apply immunotherapy against SBSN-expressing human malignancies. Consistent with the presence of SBSN expression in brains of newborn mice (Park et al., 2002), our data indicate that SBSN can be expressed under certain conditions in cells other than the suprabasal layer of the epidermis. From this perspective, SBSN may serve as a tumour neo-antigen to be exploited in anti-cancer immunotherapy, particularly for tumours resistant to standard-of-care treatment modalities such as radiation used in our present study.

Conclusions
In conclusion, our findings indicate that genotoxic stresses accompanying radiotherapy or chemotherapy can induce plastic changes of the cancer cell phenotype and increased heterogeneity of surviving cancer cell populations. We showed that survival of a subset of cancer cells in a low-adherent state depends on the concerted activity of IFNc and Erk signalling pathways activated by radiation-or chemically induced stress (Fig. 5G). Furthermore, the expression of SBSN in such low-adherent compartments of radio-and chemo-resistant cells suggests an unorthodox mechanism through which cancer cells can respond to genotoxic stress. Notably, our data highlight the dependency on SBSN as a potential vulnerability of aggressive and metastatic tumours that are commonly refractory to current standard-of-care treatments.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.  1 lM). Surviving re-adherent HeLa cells were detected on day 24 following treatment. Data are shown as mean values AE SEM, with n = 3 per group. Two-way ANOVA was used for multigroup comparisons followed by Tukey's post hoc test. The asterisk represents P < 0.05. Fig. S2. Analysis of proteome and transcriptome of 5-AC-and fIR-treated cells. (A) Genes in GOBP category: 'response to type I interferon' with their logarithm fold change values compared with control. (B) Correlation of mRNA and protein expression in irradiated (10 9 2 Gy) DU145 low-adherent cells compared with control cells based on annotations. Numbers on the x-or y-axis represent the score achieved in 2D annotation enrichment analysis. Terms with both coordinates positive and negative are upregulated/downregulated concomitantly in the proteome and transcriptome, the rest is regulated discordantly. (C) Table containing the exact terms that have both score values > 0.4 or < À0.4. Fig. S3. IFNc induced the low-adherent phenotype. (A) Loss of adhesion expressed as a ratio (fold induction) of Annexin V À /Hoechst À detached cells relative to all Annexin V À /Hoechst À cells assessed by FACS in control HS-5 cells or cells treated with IFNc (5 ngÁmL À1 ) for 3 days. (B) Phase contrast microscopic image of MCF-7 cells irradiated for 6 days with one dose of 10 Gy. Large extracellular vesicles are marked with yellow arrows. Scale bar 10 lm. (C) Immunoblotting detection of STAT1 phosphorylated on tyrosine 701 and Erk1/2 phosphorylated on threonine 202/tyrosine 204 in HeLa cells treated with IFNc (5 ngÁmL À1 ) for 72 h in the presence of MEK/ Erk inhibitor selumetinib (ERKi; 1 lM). GAPDH was used as a loading control. (D) Immunoblotting detection of STAT1 phosphorylated on tyrosine 701 and IRF1 in HeLa and MCF-7 cells treated with IFNc (5 ngÁmL À1 ) for 72 h after knockdown of IRF1 (siIRF1). Non-targeting siRNA (siNC) was used as a control. GAPDH was used as a loading control. (E) Immunoblotting detection of IRF1 in DU145 and MCF-7 cells 72 h after irradiation with one dose of 10 Gy after knockdown of IRF1 (siIRF1). Non-targeting siRNA (siNC) was used as a control. GAPDH was used as a loading control. Data are shown as mean values AE SEM, with n = 3. Twotailed unpaired Student's t-test was used for comparison between two groups for statistical significance. The asterisk represents P < 0.05. Fig. S4. Transcriptome analysis of cell populations after fIR and 5-AC exposure and SBSN detection. (A) Venn diagram of upregulated genes in low-adherent HeLa cells treated with 4 lM 5-AC for 7 days and MCF-7 and DU145 irradiated daily with 2 Gy for 10 days. Numbers represent single and common significantly up-or downregulated genes in the control (non-treated) sample. (B) Heat map showing the hierarchical clustering of LFC of genes commonly upregulated in all three low-adherent cell fractions after exposure to stress (DU145, 10 9 2 Gy; MCF-7, 10 9 2 Gy, HeLa, 5-AC). (C) RT-qPCR quantification of SBSN expression in HeLa cells after SBSN knockdown (siSBSN) treated with IFNc (5 ngÁmL À1 ) for 72 h. (D) RT-qPCR quantification of SBSN expression in HeLa and MCF-7 cells with SBSN knock down (siSBSN) treated with 4 lM 5-AC for 72 h. Immunoblotting detection of the SBSN isoforms in the cell lysates (E) and conditioned media (F) of the U373 and SK-OV-3 cells. The SBSN signal was suppressed by the SBSN siRNA (siSBSN; 48 h after the RNAi-mediated knockdown of the SBSN). Non-targeting siRNA (siNC) was used as a control. Ponceau S staining was used for a control of protein loading. Immunofluorescence detection of the SBSN in the U373 cells irradiated with a single dose of 2 Gy using LS-C162878 (G) or HPA067734 (H) antibodies. Nuclei were stained with DAPI (1 lgÁmL À1 ). Scale bar: 10 lm. (I) Quantitative FACS analysis of apoptosis using Annexin V/Hoechst staining of non-irradiated (control) or single-dose (2 Gy)-irradiated U373 cells. Non-targeting siRNA (siNC) was used as a control. Data are shown as mean values AE SEM, with n = 3. Two-way ANOVA was used for multigroup comparisons followed by Tukey's post hoc test. The asterisk represents P < 0.05. Fig. S5. Effect of the ERK1/2 pathway on the survival of cells exposed to 5-AC. (A) Immunoblotting detection of total Erk1/2 in HeLa cells treated with IFNc (5 ngÁmL À1 ) for 72 h after knockdown of Erk1 (siErk1) or Erk2 (siErk2). Non-targeting siRNA (siNC) was used as a control. GAPDH was used as a loading control. (B) Immunoblotting detection of Erk phosphorylated on threonine 202/tyrosine 204 in HeLa cells treated with 2 lM 5-AC for 72 h in the presence of MEK/Erk inhibitor selumetinib (ERKi; 1 lM). GAPDH was used as a loading control. (C) RT-qPCR quantification of SBSN expression in MCF-7 cells treated with 2 lM 5-AC for 72 h in the presence of MEK/Erk inhibitors selumetinib (ERKi; 1 lM) or U0126 (10 lM). (D) Immunoblotting detection of Erk phosphorylated on threonine 202/tyrosine 204 in MCF-7 cells treated with 2 lM 5-AC for 72 h in the presence of MEK/Erk inhibitors selumetinib (ERKi; 1 lM) or U0126 (10 lM). GAPDH was used as a loading control. (E) Immunoblotting detection of Erk1 and Erk2 in HeLa cells treated with 2 lM 5-AC for 72 h after knockdown of Erk1 (siErk1) or Erk2 (siErk2). Non-targeting siRNA (siNC) was used as a control. GAPDH was used as a loading control. Data are shown as mean values AE SEM, with n = 3 per group. Two-way ANOVA was used for multigroup comparisons followed by Tukey's post hoc test. Table S1. Transcriptome analysis of cells surviving fIR and 5-AC treatment. Excel table containing the log 2 fold-change (log 2 FC) of mRNA expression significantly deregulated in irradiated (10 9 2 Gy) DU145 and MCF-7 cells and 5-AC (7 9 4 lM)-treated HeLa cells compared to non-irradiated, non-treated control cells (1A) and results from annotation enrichment performed on clusters from the heat map (1B). Pvalues were adjusted using the Benjamini-Hochberg false discovery rate (FDR) method. Table S2. Changes of functional categories on the proteome and transcriptome level of irradiated low-adherent DU145 cells. Gene ontology (GO) biological processes (GOBP), molecular functions (GOMF), cellular compartments (GOCC), and KEGG pathways were analyzed with 1D protein (2a) and 2D protein and mRNA annotation enrichment (2b) obtained from the comparison of irradiated (10 9 2 Gy) low-adherent and non-irradiated control DU145 cells. Pvalues were adjusted using the Benjamini-Hochberg false discovery rate (FDR) method. Table S3. Description of human colon carcinoma and ovarian cancer samples. Video S1. Video of the amoeboid-like form of cell migration of low-adherent cells.