The lack of functional DNMT2/TRDMT1 gene modulates cancer cell responses during drug-induced senescence

Cellular senescence may be a side effect of chemotherapy and other anti-cancer treatments that may promote inflammation and paracrine secondary senescence in healthy tissues. DNMT2/TRDMT1 methyltransferase is implicated in the regulation of cellular lifespan and DNA damage response (DDR). In the present study, the responses to senescence inducing concentrations of doxorubicin and etoposide in different cancer cells with DNMT2/TRDMT1 gene knockout were evaluated, namely changes in the cell cycle, apoptosis, autophagy, interleukin levels, genetic stability and DDR, and 5-mC and NSUN1-6 levels. Moreover, the effect of azacytidine post-treatment was considered. Diverse responses were revealed that was based on type of cancer cells (breast and cervical cancer, osteosarcoma and glioblastoma cells) and anti-cancer drugs. DNMT2/TRDMT1 gene knockout in drug-treated glioblastoma cells resulted in decreased number of apoptotic and senescent cells, IL-8 levels and autophagy, and increased number of necrotic cells, DNA damage and affected DDR compared to drug-treated glioblastoma cells with unmodified levels of DNMT2/TRDMT1. We suggest that DNMT2/TRDMT1 gene knockout in selected experimental settings may potentiate some adverse effects associated with chemotherapy-induced senescence.


INTRODUCTION
A number of anti-cancer strategies, which are based on chemotherapy, radiotherapy and immunotherapy or the use of CDK4/6 inhibitors and epigenetic modulators may promote cellular senescence in cancer and normal cells and tissues as an adverse side effect [1,2]. Cellular senescence, a state of permanent cell cycle arrest with well characterized biochemical and molecular biomarkers, is considered to be a tumor suppressor mechanism and tissue repair and regeneration modulator [3,4]. However, in some circumstances, cellular senescence may also stimulate chronic inflammation and tumorigenesis in aged organisms that is mediated, at least in part, by persistent DNA damage signaling-induced senescence-associated secretory phenotype (SASP) [3][4][5]. Several signal transduction pathways such as NF-κB, p38 MAPK and C/EBPβ are implicated in the SASP response that result in the secretion of proinflammatory factors such as IL-6, IL-8, CXCL1, CCL2, CCL5 and matrix metalloproteinase (MMP) 3 [6][7][8][9]. SASP-mediated proinflammatory milieu may contribute to various aspects of cancer progression, namely may stimulate cancer cell proliferation, migration, invasiveness, angiogenesis and epithelialmesenchymal transition (EMT) [1,2]. Moreover, therapy-induced senescence in cancer cells is no longer considered functionally irreversible as senescent cancer cells may escape from growth arrest, resume proliferation and exhibit more invasive and migratory properties than untreated corresponding cancer cells, which in turn may drive the formation of secondary tumors or cancer relapse [1,2]. Thus, the AGING elimination of cancer-associated senescent cells both cancer cells and normal cells by senolytic drugs and the attenuation of SASP by senostatics seem a new promising approach to prevent or delay cancer recurrence and relieve frailty and multimorbidity in long-term cancer survivors [10,11]. TRDMT1, tRNA aspartic acid methyltransferase 1, (an alternative previous name DNA methyltransferase 2, DNMT2) is implicated in RNA C-5 methylation that promotes tRNA stability and protein synthesis [12][13][14][15]. The biological function of human DNMT2/TRDMT1 is not fully recognized [13], however its role has been proposed during viral infection, RNA processing and stress-mediated modulation of cell proliferation and RNA protection [16][17][18][19][20]. More recently, we have shown that DNMT2/TRDMT1 gene silencing in human fibroblasts promoted oxidative stress, genomic instability and changes in proliferation-related miRNAs that inhibited cell proliferation and stimulated cellular senescence [21]. There is limited information on DNMT2/TRDMT1-mediated effects on cancer initiation, promotion, progression and therapy [22,23]. Depletion of DNMT2/TRDMT1 gene in tumorigenic HEK293 cells resulted in modified mRNA methylation profiles and limited cell proliferation and migration [22]. It was also reported that DNMT2/TRDMT1, a writer of RNA m 5 C at sites of DNA damage, is required for efficient homologous recombination (HR) and provided resistance of U-2 OS osteosarcoma cells to radiotherapy and PARP inhibitors [23]. It was suggested that DNMT2/TRDMT1 might be considered as a novel target in cancer therapy as the loss of DNMT2/TRDMT1 sensitized cancer cells to PARP inhibitors in vitro and in vivo [23]. It is worthwhile to mention that DNMT2/TRDMT1 activity can be also pharmacologically inhibited as azacytidine (AZA) may diminish DNMT2/TRDMT1-mediated RNA methylation in human cancer cell lines [24]. However, the role of DNMT2/TRDMT1 during chemotherapyinduced senescence and related responses in cancer cells of different origin and type has not been addressed.
In the present study, we have used four different cancer cell types (breast and cervical cancer, osteosarcoma and glioblastoma cells) to analyze the effects of DNMT2/TRDMT1 gene knockout during doxorubicinand etoposide-induced senescence and the impact of AZA post-treatment. The loss of DNMT2/TRDMT1 gene affected drug-induced senescence program in glioblastoma cells that was accompanied by apoptosis resistance, DNA damage and impaired DDR and autophagic response. Detailed consequences of the lack of functional DNMT2/TRDMT1 gene during drugmediated senescence in four human cancer cell lines are presented and discussed.

DNMT2/TRDMT1 gene knockout potentiates drugmediated G2/M cell cycle arrest in breast cancer cells and glioblastoma cells
To address a universal role of DNMT2/TRDMT1 methyltransferase during anti-cancer drug-induced senescence, four different cancer cell lines were used, namely MDA-MB-231 breast cancer cells, HeLa cervical cancer cells, U-2 OS osteosarcoma cells and U-251 MG glioblastoma cells, and CRISPR/Cas9 technology was considered to knockout the DNMT2/TRDMT1 gene. Two different DNMT2 double nickase plasmids (two target-specific 20 nt guide RNAs, gRNAs) were used and two different clones with non-detectable levels of DNMT2/TRDMT1 were obtained. A representative Western blot-based analysis of DNMT2 levels is presented in Figure 1A.
Similar DNMT2/TRDMT1 gene knockout efficiency using two different target-specific gRNAs was obtained (data not shown) and further experiments were conducted using two different clones. As similar results were noticed, representative data on selected clone are presented (D-NIC cells). As control cells, cells transfected with control double nickase plasmid (nontargeting 20 nt scramble guide RNA, C-NIC cells) were used.
To promote chemotherapy-induced senescence, two anti-cancer drugs were considered, namely topoisomerase II inhibitors 35 nM doxorubicin (DOX) and 1 µM etoposide (ETOPO) and 24 h treatment. The concentrations were selected on the basis of literature data [1,25]. Fibrosarcoma, colon cancer and ovarian cancer cells were treated with DOX at the concentrations ranging from 20 to 50 nM and increased senescence-associated beta-galactosidase (SA-β-gal) activity, G1 cell cycle arrest and elevated micronuclei formation were observed [1,25]. Similar effects were noticed when fibrosarcoma cells were stimulated with 900 nM ETOPO [1,25]. In our experimental conditions, treatment with both anti-cancer drugs resulted in G2/M cell cycle arrest of the cell cycle regardless of cancer cell type (p < 0.001, Figure 1B). DNMT2/TRDMT1 gene knockout potentiated this effect in MDA-MB-231 breast cancer cells and U-251 MG glioblastoma cells (p < 0.001, Figure 1B). In contrast, a minor increase in G0/G1 cell subpopulation was noticed in ETOPOtreated HeLa cells with DNMT2/TRDMT1 gene AGING AGING knockout compared to ETOPO-treated control HeLa cells (p < 0.001, Figure 1B). The effect of 1 µM 5azacytidine (AZA, a hypomethylating agent with anticancer activity [26]) alone and in combination with anticancer drugs (post-treatment) was also studied ( Figure  1B and Supplementary Figure 1). However, AZA alone did not affect the cell cycle of cancer cells and did not potentiate the effect of anti-cancer drugs in D-NIC cells ( Figure 1B and Supplementary Figure 1). AZA and its deoxy derivative 5-aza-2'-deoxycytidine (decitabine) can be considered both effective in inhibiting DNA methylation, and 1 µM azacytidine, but not 1 µM decitabine, can also inhibit cytosine 38 methylation of tRNA Asp , a major substrate of DNMT2/TRDMT1, in different human cancer cell lines [24]. Thus, we have used relatively low sub-toxic concentration of 1 µM azacytidine to analyze the effects of AZA-mediated changes in RNA methylation on C-NIC and D-NIC cell proliferation and viability as a single and combined treatment with DOX or ETOPO (AZA post-treatment). Of course, AZA can also promote cytotoxic effects in human cancer cells, for example, apoptosis induction, when used at higher concentrations ranging from 5 to 50 µM [27].

DNMT2/TRDMT1 gene knockout limits druginduced p21-mediated senescence in glioblastoma cells
We have then asked the question of whether observed cell cycle arrest might be considered permanent. Thus, we have analyzed drug-induced cellular senescence and the effect of DNMT2/TRDMT1 gene knockout ( Figure 2). DOX and ETOPO promoted senescence-associated beta-galactosidase (SA-β-gal) activity and an increase in nuclear p21 levels in four cancer cell lines both in C-NIC and D-NIC cells (p < 0.001, Figure 2A, 2C). DNMT2/TRDMT1 gene knockout modulated senescence biomarkers only in glioblastoma cells (p < 0.001, Figure  2A, 2C). There was a decrease of 30% and 20% of SAβ-gal positive D-NIC glioblastoma cells compared to C-NIC glioblastoma cells upon DOX and ETOPO treatments, respectively (p < 0.001, Figure 2A). DNMT2/TRDMT1 gene knockout also promoted a decrease in nuclear p21 pools in DOX-and ETOPOtreated glioblastoma D-NIC cells compared to DOXand ETOPO-treated glioblastoma C-NIC cells (p < 0.001, Figure 2C). 1 µM AZA alone did not induce cellular senescence and did not potentiate the prosenescent effect of DOX and ETOPO in four cancer cell lines both in C-NIC and D-NIC cells ( Figure 2). It was previously reported that 20 µM decitabine, but not 20 µM azacytidine promoted a senescence-like phenotype in in vitro and in vivo cancer models when used for 72 h [27]. The authors concluded that two closely related DNMT inhibitor (DNMTi) nucleoside analogues azacytidine and decitabine may induce a substantially different molecular responses in cancer cells, namely azacytidine may trigger apoptosis, whereas decitabine may stimulate cellular senescence [27]. Indeed, decitabine treatment was associated with the induction of senescence program in diverse cancer cells as judged by increased SA-β-gal activity, reduced proliferation, p16/p21/p27 upregulation, SASP induction or DNA damage response [1,[28][29][30].

DNMT2/TRDMT1 gene knockout decreases drug sensitivity in glioblastoma cells
Despite the fact that we have used relatively low concentrations of anti-cancer drugs promoting senescence in cancer cells [1,25], we have then decided to evaluate DOX-and ETOPO-induced apoptosis (Annexin V staining) and the effect of DNMT2/TRDMT1 gene knockout after 24 h of drug treatments and 7 days after drug removal ( Figure 3 and Supplementary Figures 2, 3).
U-2 OS and U-251 MG D-NIC cells were more resistant to 24 h treatment with DOX compared to corresponding DOX-treated C-NIC cells (p < 0.001, Figure 3A and Supplementary Figure 2A), whereas HeLa and MDA-MB-231 D-NIC cells were more sensitive to 24 h treatment with DOX compared to corresponding DOX-treated C-NIC cells (p < 0.001, Supplementary Figure 2A). It was previously reported that a primary response to 1 µM DOX treatment for 2 h is apoptotic cell death in breast cancer cells [31]. However, the residual surviving population is subjected to accelerated senescence program that is manifested by the upregulation of p21, downregulation of cdc2 and the involvement of reactive oxygen species (ROS) regardless of p53 status [31]. The changes in phosphatidylserine externalization between ETOPOtreated C-NIC and D-NIC cells were rather mild and insignificant ( Figure 3A and Supplementary Figure 2B). One exception was ETOPO treatment in U-2 OS D-NIC cells that resulted in moderate increase in ETOPO sensitivity compared to corresponding ETOPO-treated U-2 OS C-NIC cells (p < 0.001, Supplementary Figure  2B). AZA did not promote apoptosis and did not affect DOX-and ETOPO-induced apoptotic cell death in four cancer cell lines ( Figure 3A   AGING Supplementary Figure 3A). Moreover, U-2 OS and U-251 MG D-NIC cells were more sensitive to necrosis after 7 days of ETOPO and DOX removal, respectively (p < 0.001, Figure 3B and Supplementary Figure 3). There were 12% and 10% more necrotic U-2 OS and U-251 MG D-NIC cells compared to corresponding drug-treated C-NIC cells after 7 days of ETOPO and DOX removal, respectively (p < 0.001, Figure 3B and Supplementary Figure 3). We have also analyzed senolytic activity of AZA after 7 days of drug removal and treated senescent cancer cells with AZA for 24 h. Surprisingly, DNMT2/TRDMT1 gene knockout resulted in AZA-mediated apoptosis-based senolytic activity in HeLa D-NIC cells compared to corresponding C-NIC cells upon stimulation with DOX and ETOPO (Supplementary Figure 3). In contrast, AZA-mediated senolytic activity relied on necrosis in ETOPO-treated U-2 OS D-NIC cells (Supplementary Figure 3). Thus, in these particular cellular settings, DNMT2/TRDMT1 gene knockout may be considered as a therapeutic strategy to eliminate senescent cancer cells. However, AZAmediated senolytic effects were not observed in drugtreated MDA-MB-231 and U-251 MG D-NIC cells compared to corresponding C-NIC cells ( Figure 3B and Supplementary Figure 3).

DNMT2/TRDMT1 gene knockout augments druginduced oxidative stress in glioblastoma cells
As DOX and ETOPO anti-cancer action may be mediated or accompanied by oxidative stress [31][32][33][34][35], the levels of superoxide were also examined after drug treatment and DNMT2/TRDMT1 gene knockout (Supplementary Figure 4). Indeed, DOX and ETOPO stimulations resulted in increased levels of superoxide in four cancer cell lines (Supplementary Figure 4). DOX-induced oxidative stress may stimulate apoptosis in different cancer cells [33,34], but low-dose DOX treatment (a subclinical concentration of 100 nM) may also promote oxidative stress-mediated invasiveness in U-2 OS osteosarcoma cells [36]. The free radical scavengers, glutathione (GSH) and N-acetyl cysteine (NAC) may also attenuate DOX-induced senescence response in MCF-7 and MDA-MB-231 breast cancer cells [31]. ETOPO promoted a selective non-apoptotic cell death in oncogenic Akt-transduced U-87 MG glioblastoma cells that was mediated by elevated ROS production and oxidative damage [35]. DNMT2/TRDMT1 gene knockout also potentiated druginduced oxidative stress in glioblastoma cells, whereas in DOX-and ETOPO-treated U-2 OS D-NIC cells, the levels of superoxide were decreased compared to corresponding drug-treated C-NIC cells (p < 0.001, Supplementary Figure 4). AZA did not induce oxidative stress and did not augment drug-induced superoxide production in four cancer cell lines (Supplementary Figure 4).

DNMT2/TRDMT1 gene knockout aggravates DOXinduced DSBs and impairs DDR in glioblastoma cells
As TRDMT1/DNMT2 was recently proposed to be involved in DNA damage response (DDR) in cancer cells [23], we have then analyzed the effect of DNMT2/TRDMT1 gene knockout on genetic instability ( Figure 5B, 5D). In contrast, impaired DDR was revealed in drug-treated U-251 MG D-NIC cells that was judged by decreased phosphorylation status of ATM and H2AX, limited formation of 53BP1 foci and RNA-DNA hybrid foci, and reduced levels of RAD51 and RAD52 compared to corresponding drug-treated C-NIC cells (p < 0.001, Figure 5). More recently, RNA modificationorchestrated TRDMT1-m 5 C-RAD52-RAD51 axis promoting homologous recombination (HR) at ROSinduced DSBs was described [23]. DNMT2/TRDMT1 was shown to be recruited to DNA damage sites and required for the induction of RNA m 5 C as a writer of RNA m 5 C at sites of DNA damage [23]. Loss of DNMT2/TRDMT1 affected HR, a pathway critical for DSB repair, by impaired localization of RAD51 and RAD52 to ROS-induced DNA damage [23]. RAD52 was also established to be a reader of RNA m 5 C as RAD52 showed a higher affinity for RNA-DNA hybrids containing m 5 C modified RNA than hybrids without the modification [23]. Moreover, the lack of functional DNMT2/TRDMT1 gene potentiated cancer AGING cell sensitivity to radiation and PARP inhibitors (PARPi) [23]. Except of AZA-mediated synergistic effect on DOX-induced DNA damage in U-251 MG C-NIC cells, AZA post-treatment did not significantly augment drug-stimulated DNA damage ( Figure 4A and Supplementary Figure 5). It was proposed that azacytidine has limited ability to induce DSBs and micronuclei formation compared to decitabine in different cancer models in vitro [27]. DOX and ETOPO also promoted micronuclei production, but DNMT2/TRDMT1 gene knockout did not potentiate this effect in four cancer cell lines ( Figure 4B and Supplementary Figure 5A, 5C). Moreover, XRCC1 did not seem to play an important role during DOX-and ETOPO-induced DDR in four cancer cell lines as judged by rather mild and insignificant changes in its levels. In general, XRCC1 levels decreased in our experimental settings ( Figure 5B and Supplementary Figure 5A, 5C). In DOX-treated U-251 MG cells, the levels of 53BP1 and XRCC1 were negatively correlated ( Figure 5B). DOX-treated U-251 MG D-NIC cells were characterized by decreased pools of 53BP1 and increased pools of XRCC1 compared to corresponding DOX-treated C-NIC cells (p < 0.001, Figure 5B). Indeed, 53BP1 is considered a negative regulator of alternative non-homologous end joining (NHEJ) pathway that involves XRCC1 protein [37]. As DNMT2/TRDMT1-or NSUN2-mediated methylation may modulate the interactions between tRNAs and Argonautes (Agos) to produce tRNArelated fragments (tRFs) controlling gene expression by transcriptional, post-transcriptional and translational regulation during stress conditions [38][39][40][41], we have also analyzed the levels of Ago2 during drug-induced senescence and the effect of DNMT2/TRDMT1 gene knockout (Supplementary Figure 6). DOX and ETOPO stimulations resulted in an increase in the levels of Ago2 in HeLa and MDA-MB-231 cells, but there were no differences between C-NIC and D-NIC cells (p < 0.001, Supplementary Figure 6). Thus, one can conclude that DNMT2/TRDMT1 gene knockout did not affect Ago2mediated regulatory role during stress-promoted changes in gene expression profiles.

DNMT2/TRDMT1 gene knockout diminishes druginduced autophagy in glioblastoma cells
DOX and ETOPO promoted autophagy in four cancer cell lines as judged by increased levels of cytosolic LC3 (p < 0.001, Figure 6A and Supplementary Figure 7).
However, in DOX-and ETOPO-treated U-251 MG D-NIC cells, the levels of cytosolic LC3 were significantly decreased compared to corresponding drug-treated C-NIC cells (p < 0.001, Figure 6A). Decreased levels of cytosolic LC3 were also accompanied by diminished pools of p62 and lysosome number in U-251 MG D-NIC cells upon drug stimulations (p < 0.001, Figure  6A). This may suggest that DNMT2/TRDMT1 gene knockout resulted in impaired autophagic response in glioblastoma cells. It was reported that DOX-mediated autophagy may determine cellular context-dependent cancer cell fates [31,42]. A blockage of apoptosis by pharmacological inhibitor ZVAD-fmk promoted cytotoxic autophagy in human breast cancer cells [31]. In contrast, in human osteosarcoma cells, DOX treatment stimulated cytoprotective autophagy that attenuated apoptotic cell death [42]. Genetic (ATG7 siRNA) or pharmacological (3-methyladenine) inhibition of autophagy potentiated DOX-induced apoptosis in osteosarcoma cells [42]. Except of AZAmediated increase in p62 levels in DOX-and ETOPOtreated U-251 MG D-NIC cells, the effect of AZA on autophagy parameters was limited in cancer cells with DNMT2/TRDMT1 gene knockout ( Figure 6A and Supplementary Figure 7).
Low dose DOX treatment (50 ng/ml that is equivalent to 90 nM) resulted also in hepatoma cell death through mitotic catastrophe accompanied by senescence-like phenotype [43]. DOX-induced senescence was preceded by multinucleation and downregulation of mitotic checkpoint proteins such as CENP-A, Mad2, BubR1 and Chk1 [43]. As an increase in micronuclei production and multinucleated cells (polyploid giant cells) and polyploidy may be associated with chemotherapy resistance [44][45][46] and/or a selective form of autophagy, namely nucleophagy [47][48][49], we have then analyzed drug-mediated multinucleation and the effect of DNMT2/TRDMT1 gene knockout ( Figure  6B and Supplementary Figure 7). Indeed, DOX and ETOPO treatments resulted in increased levels of multinucleated cells in four cancer cell lines (p < 0.001, Figure 6B and Supplementary Figure 7). However, except of DOX-associated decrease in the levels of multinucleated cells in HeLa D-NIC cells compared to corresponding DOX-treated C-NIC cells (p < 0.001, Supplementary Figure 7A), DNMT2/TRDMT1 gene knockout did not affect drugmediated elevation in the levels of multinucleated cells ( Figure 6B and Supplementary Figure 7). AZA also did not stimulate the occurrence of multinucleated cells ( Figure 6B and Supplementary Figure 7). It was reported that chemo-resistance (DOX, paclitaxel, docetaxel) in multinucleated breast cancer cells is achieved by ROS-HIF-1α signaling axis [46]. Multinucleated breast cancer cells are characterized by elevated levels of ROS that stabilized HIF-1α and promoted VEGF and MIF secretion, which in turn upregulated anti-apoptotic proteins contributing to chemo-resistance [46].

DNMT2/TRDMT1 gene knockout modulates SASP during drug-induced senescence in glioblastoma cells
As stress-induced premature senescence (SIPS), here DOX-and ETOPO-stimulated senescence in cancer cells, is accompanied by senescence-associated secretory phenotype (SASP) [1,8,9], we have then decided to evaluate the levels of nuclear NF-κB, a major regulator of proinflammatory response [6] and the levels of selected cytokines, namely IL-1β, IL-6 and IL-8 (Figure 7 and Supplementary Figure 8).
In general, the treatments with anti-cancer drugs did not substantially affect the levels of nuclear NF-κB ( Figure  7 and Supplementary Figure 8), thus NF-κB might not be involved in the regulation of proinflammatory secretion of senescent cancer cells in our experimental settings. Perhaps other pathways may be implicated in DOX-and ETOPO-stimulated SASP (this study) [7,8]. Indeed, DOX, ETOPO and 5-fluorouracil may activate p38 MAPK in murine macrophages and Lewis lung carcinoma (LLC1) cells [50]. Drug-induced p38 MAPK activity was associated with increased production of inflammatory cytokines IL-1β, TNF-α and IL-6 [50]. DOX and ETOPO also promoted an increase in the production of IL-1β, IL-6 and IL-8 in four cancer cell lines (p < 0.001, Figure 7A Figure 8B). IL-1 signaling may contribute to acute DOX-induced cardiotoxicity, thus its inhibition may limit cardiomyocyte toxicity in DOX-exposed patients [51]. IL-1α also increased the sensitivity of MG-63, SAOS-2 and TE-85 osteosarcoma cells to etoposide when these two agents were used simultaneously [52]. ETOPOtreated U-2 OS D-NIC osteosarcoma cells were also characterized by elevated secretion of IL-1β that was accompanied by increased sensitivity to ETOPO stimulation compared to U-2 OS C-NIC cells (p <0.001, Supplementary Figures 2, 8B). IL-1 signaling may also modulate tumor microenvironment and promote cancer initiation and progression [53]. Moreover, IL-1mediated activation of anti-apoptotic signaling pathways and senescence may contribute to resistance to targeted therapies [53]. IL-1α and IL-1β signals through IL-1R may promote the SASP response in a cooperative manner that may trigger some protumorigenic effects [54]. Indeed, IL-1α inactivation impaired tumor progression and immune cell infiltration without affecting cell cycle arrest in a mouse model of pancreatic cancer [54]. AZA did not promote a proinflammatory response in four cancer cell lines (Figure 7 and Supplementary Figure 8).

DNMT2/TRDMT1 gene knockout potentiates druginduced increase in the levels of NSUN6 in glioblastoma cells
We were then interested if DNMT2/TRDMT1 gene knockout may modulate the levels of other RNA methyltransferases, namely the NOL1/NOP2/SUN domain (NSUN) family proteins NSUN1, NSUN2, NSUN3, NSUN4, NSUN5 and NSUN6 (Figure 8 and Supplementary Figure 9). Surprisingly, except of the levels of NSUN2, the pools of NSUN proteins were elevated during DOX-and ETOPO-mediated senescence in four cancer cell lines (p < 0.001, Figure 8A and Supplementary Figure 9). In general, DNMT2/TRDMT1 gene knockout did not augment the levels of NSUN proteins after drug treatment (Figure 8 and Supplementary Figure 9). One exception was the elevation of NSUN6 levels in DOXand ETOPO-treated U-251 MG D-NIC cells compared to corresponding drug-treated C-NIC cells (p < 0.05 and p < 0.01, Figure 8A). AZA did not affect the levels of NSUN proteins in four cancer cell lines (Figure 8 and Supplementary Figure 9). There are conflicting reports on the complex role(s) of NSUN proteins in cancer biology and cellular senescence [55][56][57]. For example, NSUN2 may both delay and stimulate cellular senescence that may reflect different cellular contexts [56,57]. NSUN proteins may also exert both cancer promoting and inhibiting effects [58][59][60][61]. NSUN6 may suppress the proliferation and development of liver and pancreatic cancer cells [60,61]. However, in our experimental conditions, increased levels of NSUN6 in DOX-and ETOPO-treated U-251 MG D-NIC cells were accompanied by apoptosis resistance compared to corresponding C-NIC cells (Figures 3, 8).

DNMT2/TRDMT1 gene knockout lowers the cytoplasmic pools of 5-methylcytosine during druginduced senescence in glioblastoma cells
We have then analyzed if the lack of active DNMT2/TRDMT1 gene may affect an epigenetic mark, namely the levels of nuclear and cytoplasmic fractions of 5-methylcytosine. DOX and ETOPO treatments AGING AGING resulted in increased levels of both nuclear and cytoplasmic pools of 5-methylcytosine in four cancer cell lines (Supplementary Figure 10). DNMT2/TRDMT1 gene knockout-associated decrease in cytoplasmic fraction of 5-methylcytosine was only observed in DOX-and ETOPO-treated U-251 MG D-NIC cells, ETOPO-treated HeLa D-NIC cells and DOX-treated U-2 OS D-NIC cells compared to corresponding drugtreated C-NIC cells (p < 0.05 and p < 0.001, Supplementary Figure 10). As the levels of NSUN proteins were increased during DOX-and ETOPOinduced senescence in four cancer cell lines (p < 0.001, Figure 8A and Supplementary Figure 9), one can speculate that limited impact of DNMT2/TRDMT1 gene knockout on the cytoplasmic pools of 5-methylcytosine is based on a compensation effect of elevated levels of NSUN proteins. However, we analyzed immunofluorescence-based changes in global 5methylcytosine that did not allow for the discrimination between methylation patterns of different types of RNAs. AZA did not substantially affect the levels of nuclear and cytoplasmic 5-methylcytosine in four cancer cell lines (Supplementary Figure 10).
In conclusion, we have shown for the first time that DNMT2/TRDMT1 gene knockout affected DOXand ETOPO-mediated senescence program in glioblastoma cells as judged by modulated SIPS response (decreased number of SA-beta-gal-positive cells and nuclear p21 pools), apoptosis resistance, increased ROS levels, elevated DNA damage and impaired RNA-mediated DDR, limited autophagic response, modulated SASP and affected NSUN levels ( Figure 9).
Thus, the lack of functional DNMT2/TRDMT1 gene in glioblastoma cells may potentiate some adverse side effects associated with chemotherapy-induced senescence. A summary of obtained results using four different cancer cell lines, namely HeLa, MDA-MB-231, U-2 OS and U-251 MG cells is also presented in Table 1.

Anti-cancer drug treatment
Three anti-cancer drugs, namely doxorubicin hydrochloride (DOX, 44583), etoposide (ETOPO, E1383) and 5-azacytidine (AZA, A2385) were obtained from Merck KGaA (Darmstadt, Germany). To induce drug-mediated senescence, C-NIC and D-NIC cells were treated with 35 nM DOX or 1 µM ETOPO for 24 h and then cultured without drugs for 7 days. Moreover, the action of AZA, a hypomethylating agent, was also studied in D-NIC cells lacking active DNMT2/TRDMT1 gene to analyze combined effects of the inhibition of RNA and DNA methylation. Briefly, C-NIC and D-NIC cells were treated with 1 µM AZA for 24 h after 7 days of DOX or ETOPO removal. Moreover, DOX-treated or ETOPO-treated cells for 24 h were stimulated with 1 µM AZA for additional 24 h and culture was then terminated. If not indicated otherwise, the parameters were investigated after 7 days of drug removal (DOX or ETOPO) and 24 h treatment with AZA.

Cell cycle analysis
C-NIC and D-NIC cells were treated with 35 nM DOX or 1 µM ETOPO for 24 h and/or with 1 µM AZA for additional 24 h and drug-mediated changes in the phases of the cell cycle (G0/G1, S, and G2/M) were then revealed using a Muse ® Cell Analyzer and a Muse ® Cell Cycle Assay Kit according to the manufacturer's instructions (Luminex Corporation, Austin, TX, USA).  [62]. Two subpopulations were considered, namely, superoxidepositive and superoxide-negative cells, and representative histograms are presented (%).

Immunofluorescence and GFP-based imaging
Cells were fixed and immuno-stained as comprehensively described elsewhere [63]. The following primary and secondary antibodies were considered, namely, anti-p21 (

Genetic instability, DNA damage and DNA damage response (DDR)
Micronuclei (MN) formation and multinucleation events were evaluated using Hoechst 33342 staining and digital cell images were captured using a laserbased confocal imaging and HCA system IN Cell Analyzer 6500 HS (Cytiva, Marlborough, MA, USA). Quantitative analysis was conducted using IN Carta software (Cytiva, Marlborough, MA, USA). Micronuclei production and multinucleation events were expressed as %.

Statistical analysis
The results represent the mean ± SD from at least three independent experiments. Box and whisker plots with median, lowest, and highest values were also used. Differences between control samples and treated samples were analyzed using one-way ANOVA and Dunnett's multiple comparison test, whereas differences between C-NIC cells and D-NIC cells were revealed using one-way ANOVA and Tukey's multiple comparison test. Statistical significance was evaluated using GraphPad Prism 5. P-values of less than 0.05 were considered significant.

CONFLICTS OF INTEREST
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. . Box and whisker plots are shown, n = 3, *** p < 0.001, ** p < 0.01, * p < 0.05 compared to CTR (ANOVA and Dunnett's a posteriori test), ### p < 0.001, ## p < 0.01, # p < 0.05 compared to C-NIC cells at the same culture conditions (ANOVA and Tukey's a posteriori test). CTR, control conditions; DOX, doxorubicin treatment; ETOPO, etoposide treatment; AZA, azacytidine treatment; MIX, azacytidine post-treatment; C-NIC, control cells with unmodified levels of DNMT2/TRDMT1 containing control plasmid; D-NIC, cells with DNMT2/TRDMT1 gene knockout containing dedicated DNMT2 double nickase plasmid.