ATF3 reduces migration capacity by regulation of matrix metalloproteinases via NFκB and STAT3 inhibition in glioblastoma

Glioblastoma is associated with poor survival and a high recurrence rate in patients due to inevitable uncontrolled infiltrative tumor growth. The elucidation of the molecular mechanisms may offer opportunities to prevent relapses. In this study we investigated the role of the activating transcription factor 3 (ATF3) in migration of GBM cells in vitro. RNA microarray revealed that gene expression of ATF3 is induced by a variety of chemotherapeutics and experimental agents such as the nitric oxide donor JS-K (O2-(2,4-dinitrophenyl) 1-[(4-ethoxycarbonyl)piperazin-1-yl]diazen-1-ium-1,2-diolate). We found NFκB and STAT3 to be downstream targets inhibited by overexpression of ATF3. We demonstrate that ATF3 is directly involved in the regulation of matrix metalloproteinase expression and activation. Overexpression of ATF3 therefore leads to a significantly reduced migration capacity and induction of tissue inhibitors of matrix metalloproteinases. Our study for the first time identifies ATF3 as a potential novel therapeutic target in glioblastoma.


INTRODUCTION
Glioblastoma multiforme (GBM) is the most frequent and malignant brain tumor with few therapeutic options and a poor clinical outcome. The understanding of the mechanisms that drive tumor progression plays a central role in the current GBM research, while elevated invasive capacity and multifocal growth are of great importance from a clinical point of view. Understanding the molecular regulatory mechanisms is therefore fundamental for the development of novel anti-tumor therapies to achieve tumor elimination. The ability to degrade extracellular proteins is essential for any individual cell to interact properly with its immediate surroundings. For multicellular organisms it is a prerequisite for development and chemotaxis. 1 In physiological conditions, the extracellular matrix (ECM) is involved in wound healing and organogenesis; in pathological conditions it is significantly involved in local tumor progression and metastasis in cancer. 2,3 Migration and metastasis are complex interactions between cells and the ECM. There are various groups of proteolytic enzymes or proteases involved in matrix degradation, but the matrix metalloproteinase (MMP) group of enzymes is the most important in the context of tumor invasion and metastasis. 2,4 The MMPs are zinc and calcium-dependent endopeptidases that modify signaling pathways and cell functions under physiological and pathological conditions. 5 MMPs are involved in migration, inflammation, proliferation, apoptosis and differentiation and are therefore latently synthesized. 5,6 The substrates of MMPs are specific proteins of the ECM like collagen, fibrin, integrins and receptors. 2,7 Various studies demonstrated a correlation between expression of MMPs and the malignancy and invasive capacity of GBM. 8 Certain MMPs such as MMP1 and MMP7 are associated with a poor prognosis in cancer. 5,9,10 MMPs are secreted in the ECM as inactive zymogens interacting with tissue inhibitors of metalloproteinases (TIMP). 4 TIMPs bind to the catalytic domain of MMPs and can inhibit the 'cysteine-switch', which disrupts the interaction between a cysteine in the prodomain and the Zn 2+ ion in the active site. After the cleavage of the complex the prodomain of the MMP is released for activation. [11][12][13] In general, each TIMP can form complexes with each MMP but there are preferential interactions with higher affinity between some of them. 14 The 28 MMPs described in human are classified by the prodomain, amino acid sequence and substrate specificity. 2 The AP-1 binding site has been considered to play an important role in the transcriptional activation of the MMP promoters. Activating transcription factor 3 (ATF3) is a stress-induced and adaptiveresponsive gene that encodes a member of the ATF/CREB super family of transcription. 15 The basal level of ATF3 is very low but can rapidly increase under stress. ATF3 is activated by forming a homodimer or heterodimers with ATFs or proteins of the CREB/ CREM and AP-1 families, which contain consensus sequences. 16 ATF3 functions are variable in a context-dependent manner, therefore research efforts to delineate its function are difficult. DNA damage, non-steroidal anti-inflammatory drugs, progesterone and even mere cellular stress can induce ATF3 expression and activation. 17,18 Recent studies report that ATF3 has neuroprotective functions, for example, through downregulation of inflammatory cytokines or prevention of apoptosis. 16,17 Overexpression of ATF3 in lung cancer cells leads to reduced metastasis, whereas knockdown of ATF3 results in a higher migration capacity with increased levels of actin filaments and correlates with a shorter overall survival in esophageal squamous cell carcinoma. 19,20 Since ATF3 is known to be stress inducible, it is conceivable that it is regulated by factors involved in stress response and inflammation such as nitric oxide (NO). NO is a free radical with diverse functions in immunoreactions and neurotransmission. 21 Endogenous NO is generated by eNOS and nNOS under physiological conditions in the brain and may act as a cytoprotective and proliferative element in neuronal cells. 22,23 One explanation for this cytoprotection is the ability of NO to mediate cGMP generation and therefore the differentiation of precursor cells and prevention of apoptosis after stimulation which could be confirmed in myogenic cells. [24][25][26] To stabilize the reactive and diffusible NO and to facilitate delivery of therapeutic NO doses, a prodrug was developed for in vitro and in vivo usage. The prodrug JS-K (O2-(2,4-dinitrophenyl) 1-[(4-ethoxycarbonyl)piperazin-1-yl]diazen-1ium-1,2-diolate) is a diazeniumdiolate that releases NO after enzymatic metabolization by glutathione S-transferases (GSTs). 27 In previous studies we could show the specific release of NO by JS-K in GST-overexpressing GBM cells affecting their proliferation activity and viability in a dose-and time-dependent manner. 28 Furthermore, we could show that cells exposed to high concentrations of NO undergo mitotic catastrophe resulting in necrosis. 29 Nitric oxide (NO) in low concentrations inhibits induction of cell death by S-nitrosylation of signaling molecules such as NFκB, caspases and transcription factors. [30][31][32] There is some evidence that NO released by NO donors at non-cytotoxic doses affects the migratory capacity of tumor cells. 33,34 The main objective of this study was to investigate the molecular mechanisms of migration induced by low dose NO released from the diazeniumdiolate JS-K in glioma cells in vitro. ATF3 was found to be upregulated by NO in RNA microarray. In this study we provide evidence that overexpression of ATF3 correlates with expression and activation of MMPs and TIMPs in vitro. The reduced activation of MMPs leads to restricted migration capacity of established and primary GBM cells in vitro. In addition, we demonstrate a positive correlation between ATF3 and phosphorylation of STAT3 and the inhibition of nuclear translocation of NFκB in glioblastoma cells. We show that the transcription factor ATF3 directly regulates gene expression of nfκb, stat3 and klf6 that are involved in tumor progression and invasion. Therefore, we propose that ATF3 could serve as prognostic marker for migration and metastatic disease in glioblastoma.

RESULTS
Nitric oxide reduces migration capacity in a time-and dose-dependent fashion To investigate the influence of JS-K on migration capacity of glioma cells, we performed a wound closure assay over 96 h for U87 and primary IC cells ( Figure 1). Starting at 24 h, the ability to close the migration gap is time-dependently reduced in both cells lines. In U87 cells, this significantly reduced effect can be observed at a concentration of 2 μM JS-K (Figure 1a), whereas in primary IC cells it emerged at 1 μM (Figure 1b). Up to 3 μM JS-K progressively inhibited migration in a dose-dependent manner.
Migration capacity is not influenced by reduction of viability and proliferation One explanation for cells not closing a wound can be increased cell death elicited by the treatment or the inhibition of proliferation. To exclude these factors caused by JS-K we investigated cell viability by MTT assay over 96 h as well as the proliferation rate by BrdU incorporation assay (Figures 1c-f). Less than 20% of U87 cells were killed at the highest dose of 3 μM after 96 h (Figure 1c). Viability of primary IC cells was not affected by JS-K up to 72 h after treatment (Figure 1d). At 96 h, the viability decreased to 68% when using 3 μM JS-K. In contrast, no influence of JS-K on proliferation was detected at the tested doses at all Microarray analysis reveals activating transcription factor 3 to be upregulated by NO To identify the gene targets of NO, we determined the temporal gene expression profile of U87 cells exposed to NO for 48 h using RNA microarray. We used U87 cells as a test cell line because this cell line is well characterized for many years as a standard for biochemical and genetic experiments in glioblastoma research. The data set was analyzed for upregulated and downregulated genes more than 2-fold in a statistically significant manner (P ≤ 0.05). Approximately 10 genes were found to be upregulated (Table 1) and further analyzed by qRT-PCR to validate the array (data not shown). Activation transcription factor 3 (ATF3) was found to be upregulated 2.51-fold in the microarray. Primary IC cells and U87 cells revealed dose-dependent upregulation of expression in qRT-PCR after 48 h exposure to JS-K at concentrations up to 15 μM ( Figure 2). U87 cells showed a 15-fold expression compared to controls at the highest concentration of 15 μM (Figure 2a). Primary IC cells had the highest peak of 5.8-fold expression when using 10 μM JS-K ( Figure 2b). ATF3 is known for its key role in suppression of metastasis and invasion. Therefore we chose migration for further investigation of the role of ATF3 in glioblastoma. To demonstrate the relevance of ATF3 we examined ATF3 RNA expression in response to treatment with various anticancer agents (Figure 2c). Epidermal growth factor (EGF) and hydrogen peroxide (H 2 O 2 ) which is known to induce necrosis, did not induce the expression of ATF3 more than 2-fold. Temozolomide (TMZ), a common chemotherapeutic agent used in GBM therapy, did not even induce a twofold upregulation in contrast to Etoposide (ETO), also used for GBM therapy, leading to a 16-fold expression of ATF3. Paclitaxel (Taxol), a chemotherapeutic for a variety of tumors, showed a 9.5-fold expression and the antiinflammatory acetylsalicylic acid (ASA) a 3.4-fold expression of ATF3. This highlights the considerable importance of research into ATF3 in glioblastoma.
Overexpression of ATF3 reduces migration capacity and proliferation in vitro Since we observed the overexpression of ATF3 in GBM cells induced by NO as well as different anticancer agents, we overexpressed and knocked down ATF3 stably by lentiviral infection in U87 cells. The efficacy of the knockdown (shATF3) and the overexpression (ATF3), done with their controls, pGIPZ and pLOC, are shown in Figures 3a and b. Knockdown, overexpression and treatment with JS-K are normalized to the particular control set to one. The controls still react to JS-K as the uninfected cells (U87). The knockdown of ATF3 led to a decreased ATF3-expression to 25% compared to control. The overexpression showed RNA levels of more than 650-fold compared to control and could be still increased by JS-K. To demonstrate the effect of ATF3 on migration of U87 cells we repeated the wound closure assay. Cells infected with pLOC, pGIPZ and shATF3 did not show any difference in migration capacity compared to uninfected U87 cells over the 120 h observation period (Figures 4a and b). However, ATF3overexpressing cells migrated significantly slower into the gap (Figures 4a and b). While all cells not overexpressing ATF3 closed the gap within 120 h, the ATF3 overexpressing cells still exhibited a gap of about 1 mm, which accounts for 50% of the initial space ( Figure 4b). To reveal an association between reduced migration and reduced viability of ATF3 overexpressing cells, we performed Matrigel invasion assay. Compared to controls, knockdown of ATF3 did not lead to altered invasive capability, whereas the overexpression of ATF3 decreased invasion of the cells significantly (P = 0.00004) to 32% (Figures 4c and d). Since ATF3 is upregulated by NO, we expected a different behavior to JS-K in cell death induction, especially of shATF3. Cell lines showed no difference in viability when exposed to JS-K up to 25 μM for 48 h in the MTT assay ( Figure 4e). The relative cell proliferation measured by the BrdU incorporation assay indicated a significant reduction in ATF3 cells after 72 h (Figure 4f). Both controls and shATF3 showed the same proliferation rate compared to uninfected U87.
Overexpression of ATF3 does not regulate mRNA expression of migration-related genes but prevents translocation of NFκB and STAT3 phosphorylation The transcription factor ATF3 is known to regulate the expression of a number of genes. To investigate the relevance of ATF3 for migration we performed qRT-PCR for p53, nfκb1, stat3, p38α and klf6. Figure 5a shows the regulation of these genes in ATF3   were upregulated more than 2-fold in response to overexpression of ATF3 but not p53, nfκb1 and p38α. However, even if stat3 mRNA expression was upregulated, Western blot revealed that STAT3 is no longer phosphorylated with ATF3 overexpression (Figure 6b). When further treated with up to 5 μM JS-K, we observed a threefold upregulation of nfκb1 mRNA expression after 48 h (Figure 5b) but no difference in p53 (Figure 5c), stat3 (Figure 5d), p38α ( Figure 5e) and klf6 (Figure 5f). Since the NFκB pathway plays an important role for migration and invasion of tumor cells, we further investigated the activation of NFκB. Immunocytochemistry in cells treated with 5 μM JS-K for 48 h showed no translocation of NFκB (p65) into the nucleus caused by NO in either cell line, control or overexpressing ATF3 (Figure 6a). p65 could be detected in the cytoplasm with a much higher protein level in NO-treated cells compared to controls. Treatment with TNFα is well known to induce translocation of p65 and was used in this experiment as a positive control of translocation. Control cells translocate p65 after stimulation with TNFα for 6 h but not cells overexpressing ATF3 (Figure 6a). NFκB (p65) remained in the cytoplasm and was barely detectable in the nuclei.
Overexpression of ATF3 regulates protein level of matrix metalloproteinases and tissue inhibitors of metalloproteinases 3 Matrix metalloproteinases (MMPs) stimulate tumor metastasis by the degrading extracellular matrix. In this study we found no visible regulation of RNA expression in PCR caused by NO in MMP2, MMP7 and MMP9, nor in their inhibitors TIMP1 and TIMP3 in either U87 (Figure 7a) or primary IC cell lines (Figure 7b). TIMP2 was slightly reduced by increasing concentrations of JS-K in U87, whereas TIMP4 was dose-dependently enhanced by NO (Figure 7a). In U87 cells with ATF3 knockdown we did not observe any difference in expression of MMP1-3, MMP7 or MMP9 compared to pGIPZ or uninfected controls (Figure 7c). In contrast, cells overexpressing ATF3 exhibited no RNA expression of MMP7 compared to pLOC and uninfected cells. TIMP3 was significantly overexpressed in ATF3 overexpressing cells. Since MMP2, MMP7 and MMP9 are important for the degradation of the tumor matrix, we investigated their protein levels by Western blot and the activity of MMPs by zymography (Figure 8). Western blot of the conditioned media obtained from U87 and primary IC cells showed a reduced protein level of MMP2, MMP7 and MMP9 with increasing concentrations of JS-K up to 3 μM and an enhanced expression of TIMP3 at 1 and 2 μM in U87 and primary IC cells (Figures 8a and b). The whole cell lysate of U87 cells exhibited an almost equal expression level of MMP2, MMP7 and TIMP3 but a dose-dependently decreasing level of MMP9 ( Figure 8c). However, whole cell lysates of primary IC cells revealed decreasing protein levels of MMP2, MMP7 and MMP9 with low dose JS-K and no change in TIMP3 expression (Figure 8d). The protein level of MMP2, MMP7 and MMP9 in the supernatant of ATF3 overexpressing cells was considerably reduced compared to pLOC controls as well as the protein level of TIMP3 (Figure 8e). Western blot of the whole cell lysate of ATF3 knockdown and overexpression indicated a total deficiency of MMP2 and MMP9 in ATF3 overexpressing cells and a lower protein level of MMP7 (Figure 8f). In contrast, the inhibitor TIMP3 was overexpressed in ATF3 cells compared to controls and knockdown (Figure 8f). Zymography analysis of the conditioned media revealed no change in the activation of MMP2 in U87 exposed to NO (Figure 8a), whereas IC cells exhibitied less activation when treated with 3 μM JS-K (Figure 8b). However, ATF3 overexpression reduced the level of active MMP2 in the supernatant of the cells compared to ATF3 knockdown and controls (Figure 8c). MMP7 and MMP9 could not be detected in gelatin or casein gels (data not shown).

DISCUSSION
Exploration of molecular mechanism in GBMs is of great importance for a full understanding of the biology of this highly malignant cancer and the identification of potential therapeutic targets. Recent research suggest that NO is a putative adjuvant in anti-tumor therapy but some mechanisms remain unclear so far. [35][36][37] To obtain a new insight into the role of NO in glioblastoma biology and treatment, we investigated its role on gene targets and molecular mechanisms of cell migration. NO is known for its antiproliferative and cytotoxic effects in glioblastoma cells at high concentrations and its impact on migration, invasion and angiogenesis in various tumor cells at low doses. 28,29,38 JS-K is an arylating agent designed to release nitric oxide upon reaction with glutathione in the presence of glutathione S-transferase, an enzyme overexpressed in cancer cells. 25,28,39 Thus, JS-K application in GBM cells allows a cell typespecific intracellular NO release. The impact of high dose JS-K was described by our group as dose-and time-dependent concerning the inhibition of proliferation and induction of cell death. 28,29,40 NO induces DNA double-strand breaks and post-translational modifications of proteins, such as S-nitrosylation of NFκB resulting in activation of caspase-dependent and caspase-independent pathways. 31,32,38 In previous studies we observerd the strongest cellular effects in glioblastoma cells after 48 h of JS-K treatment. In this study we found the transcription factor ATF3 to be upregulated by 15 μM JS-K by RNA microarray. We provide ATF3 reduces migration capacity in glioblastoma J Guenzle et al evidence in established and primary cell lines that ATF3 is doseand time-dependently regulated by NO in concentrations between 1 and 15 μM. In cell culture, we could observe that the migration capacity was reduced under treatment with JS-K. While cell viability and proliferation are not affected by up to 3 μM JS-K, cells migrated slower in a dose-and time-dependent fashion. Cell migration and invasion are mainly driven by MMPs in vivo. 6,9 Recent research demonstrates that AP-1 protein complexes are involved in expression and activation of MMPs. 4,41,42 Since ATF3 is known to form heterodimers with AP-1 proteins, we hypothesize that ATF3 is a direct regulator of MMP expression and activation. Cells secrete MMPs as zymogens into the ECM where they must be activated through enzymatic cleavage. By western blot we could demonstrate that NO induces protein expression and secretion of MMP2, 7 and 9 dose dependently in the supernatant of the cultured cells. This dose-dependent capacity indicates that NO is not activating a long signaling cascade but acts in a more direct way on the gene expression. The reduced expression of MMP2, 7 Proliferation was normalized to T0 (start of experiment) set to 1. Asterisks (*Po 0.05, **P o0.01, ***Po0.001) indicate significance to controls or T0. and 9 in cell lysates as well as in the supernatant of ATF3 overexpressing cells demonstrates the direct impact of the transcription factor on proteins involved in migration and invasion. Since RNA expression of MMPs is not affected by the dose-dependent regulation of NO and ATF3, we conclude that the principle of regulation of MMPs must occur on a post-translational level. The activation of MMPs can be investigated by zymography. 4 In the conditioned media of cells exposed to NO we observed a dose-dependent reduction of activity of MMP2 in both cell lines, which implies a reduced cleavage of the ECM and therefore less migration and invasion capacity. 3 Activation of MMP2 is significantly reduced when ATF3 is overexpressed but we could not observe any difference when ATF3 is knocked down. The efficiency of the ATF3 knockdown in U87 cells was at about 75%. qRT-PCR revealed that ATF3 is basically less expressed in established and primary cell lines. However, the remaining 25% of expressed ATF3 can still be sufficient to regulate invasive processes. Due to the same specific substrate of the gelatinases, we expected to induce activation of MMP2 as well as MMP9, but we were not able to detect active MMP9 in the gelatin gels under a variety of conditions. The activation of the matrilysin MMP7 can be monitored by PAGE containing casein. 5 Both in the supernatant of cells exposed to NO and in ATF3 overexpressing cells only a weak but equal signal was detectable. We could not observe any difference induced by NO or ATF3 (data not shown). The zymogens are released from the cytoplasm of tumor cells where they are inhibited by TIMPs. 13,14 The excessive cleavage of the ECM associated with an imbalance of the MMPs/TIMPs ratio has been correlated with the invasive potential of brain tumor cells. 7,43 We could observe by PCR that TIMP3 is highly expressed in cells overexpressing ATF3 but not following exposure to JS-K up to 3 μM. We conclude that the regulation of the inhibitors is not directly affected by NO and may need stronger promotor induction. However, we could detect that protein expression of TIMP3 also increased by NO in both the cell lysate and the supernatant of the cells. In ATF3 overexpressing cells, we could detect overexpressed TIMP3 in the lysate but not in the supernatant. This leads us to the assumption that TIMP3 is released from MMP binding during secretion from the cells when ATF3 is overexpressed. Migration and invasion are regulated by several signaling pathways. NFκB1, KLF6, p53 and p38α are known to be involved in cell migration and invasion. [44][45][46][47][48] Gene expression of nfκb1 is not directly affected by ATF3 as observed by qRT-PCR. However, in cells overexpressing ATF3, the nuclear translocation of NFκB could no longer be induced by TNFα. In contrast, NO induces nfκb1 gene expression and protein expression of p65 even in ATF3 Figure 5. Relative gene expression of ATF3 overexpressing cells was determined by qRT-PCR. Expression of p53, nfκb1, stat3, p38α and klf6 in ATF3 overexpressing U87 cells was normalized to controls of uninfected U87 cells (a). Treatment with 1-5 μM JS-K for 48 h was performed in ATF3 overexpressing cells and pLOC controls (b-f). Relative gene expression of nfκb1 (b), p53 (c), stat3 (d), p38α (e) and klf6 (f) was investigated by qRT-PCR. Expression of pLOC was subtracted from expression of ATF3 overexpression to demonstrate the exclusive effect of ATF3 on target gene mRNA. Data were normalized to untreated controls ± S.D. of triplicate.
ATF3 reduces migration capacity in glioblastoma J Guenzle et al overexpressing cells but it is not translocated into the nuclei by NO. This indicates that the role of NO is not based on the same signaling pathways as ATF3. Marshall et al. and other groups found in lung carcinoma that NFκB is nitrosylated and therefore inhibited by NO in vitro. 31,49 In previous studies the S-nitrosylation of proteins induced by JS-K was well investigated and we conclude that nuclear translocation of NFκB is inhibited by direct interaction with NO. 50,51 The gene expression of tumor protein p53 plays an important role in migration research. Yan et al. found that ATF3 activates p53 in colon carcinoma cells. 52 In our study, p53 is controlled neither by ATF3 nor by NO in low doses.
Translocation of p53 into the nuclei is also not affected by ATF3 or NO (data not shown). Xu et al. 53 found that KLF6 was upregulated by ATF3 in glomerular mesangial cells to induce apoptosis. Ahronian et al. 46 identified KLF6 as a potent regulator of migration in hepatocellular carcinoma cells. Gene expression of klf6 is significantly upregulated by overexpression of ATF3 in glioblastoma cells. Furthermore, klf6 is not upregulated in cells exposed to NO. We therefore suggest other pathways to be involved in the regulation of KLF6, although Xu et al. 53 hypothesize a direct relation between ATF3 and KLF6. To investigate the involvement of ATF3 in the observed alterations after NO exposure, we No further upregulation in gene expression by NO in ATF3 overexpressing cells points out that ATf3 signaling is the major pathway triggered by NO. Many groups found STAT3 to be constitutively phosphorylated and activated in tumor cells, and inhibitors are already applied in tumor-immunology in patients. [54][55][56] The phosphorylation status of STAT3 can give an indication of the malignancy and the proliferation of tumor cells. Downregulation of STAT3 and inhibition of phosphorylation is supposed to reduce migration and invasion capacity in glioma cells. 56,57 In our study, we found gene expression of stat3 upregulated by ATF3. Gene expression was not influenced by NO in ATF3 overexpressing cells. Even though stat3 is upregulated by ATF3 it is no longer phosphorylated in ATF3 overexpressing cells as shown by Western blot. Only active STAT3 can interact with NFκB and induce MMP expression. 54,58 These data indicate that ATF3 can directly elicit MMP expression and activation. Furthermore, it recruits wellinvestigated signaling pathways to induce MMP expression and reduce migration and invasion in glioblastoma cells. Most of the MMPs are not detectable in healthy brain tissue and the presence of MMPs suggest a direct correlation to the malignancy of the tumor and the clinical outcome of the patient. 8,9,59 In the past decade all clinical studies for MMP inhibition and immunotherapy failed and showed no benefit on migration and tumor progression. 60 Therefore we conclude that ATF3 may be a novel target of migration in glioblastoma and should be further investigated.

RNA interference and lentiviral transduction
Human shRNAs on lentiviral pGIPZ plasmid targeting ATF3 (V2LHS_172635, V3LHS_405369, V3LHS_352238, V3LHS_405368, V3LHS_352240, V3LHS_40 5370), overexpression of human ATF3 on lentiviral Precision LentiORF Collection plasmids (pLOC, pLOHS_100010432, pLOHS_100005353)) and negative controls (pGIPZ, pLOC) were transfected into HEK293T cells by cotransfection of lentiviral plasmids pCMV-MD2G and psPAX2 using HiPerFect transfection reagent (Qiagen, Hilden, Germany). Virus containing supernatant was collected and used for infection of U87 cells. shATF3 pLOHS_100010432 was found to knockdown ATF3 expression most efficiently and was used for the experiments as well as V3LHS_405370 for overexpression of ATF3. The functionality of the shRNAs and overexpression plasmids were validated by qRT-PCR analysis. Transduced cells were selected for puromycin resistance before further analysis.

Migration assay
Cells were seeded in 24-well plates with silicon bars to create a gap of about 2 mm. Non-infected cells were treated with JS-K up to 3 μM while cells overexpressing ATF3 or knockdown cells were not treated. After removing the bars cells started migrating from the edge of the wound and repopulated the gap area. The time required for wound closure was measured by microscopy (Zeiss Axiovert, Oberkochen, Germany) and documented up to 120 h.

MTT-assay
Cell viability was determined by MTT Assay. Cells (10 4 ) were grown in 96well plates with complete DMEM and treated with 1-25 μM JS-K for up to 96 h. MTT Assay was performed as described before. 28 Absorbance at 570 nm was measured with Tecan Infinite200 (Tecan, Männedorf, Switzerland). Percentages were calculated relative to the viability of untreated controls set to 100%.
Cell proliferation assay Cell proliferation was monitored by 5′-bromodeoxyuridine (BrdU) incorporation assay (Roche Diagnostics, Mannheim, Germany). Three thousand cells were seeded in 96-well plates and treated with JS-K up to 3 μM. Cells overexpressing ATF3 or knockdowns were not treated. Cells were stained with BrdU following the manufacturer's instructions. The percentage of cells exhibiting genomic BrdU incorporation was measured by absorbance at 370 nm with Tecan Infinite200 (Tecan, Männedorf, Switzerland). Percentages were calculated relative to the proliferation of untreated controls or T0.

Microarray
Gene expression changes were investigated by Microarray using the Illumina HumanHT-12 Expression BeadChip (Illumina Inc., San Diego, CA, USA) for analysis of 31 000 genes. The array was performed at the German Cancer Research Center (DKFZ), Heidelberg, Germany. RNA samples were purified from U87 treated with 15 μM JS-K for 48 h as well as from untreated controls. Gene up-and downregulation caused by JS-K was evaluated. Microarray was done in triplicate.