Poly(ADP-ribose) Polymerase 1 Mediates Rab5 Inactivation after DNA Damage

Parthanatos is programmed cell death mediated by poly(ADP-ribose) polymerase 1 (PARP1) after DNA damage. PARP1 acts by catalyzing the transfer of poly(ADP-ribose) (PAR) polymers to various nuclear proteins. PAR is subsequently cleaved, generating protein-free PAR polymers, which are translocated to the cytoplasm where they associate with cytoplasmic and mitochondrial proteins, altering their functions and leading to cell death. Proteomic studies revealed that several proteins involved in endocytosis bind PAR after PARP1 activation, suggesting endocytosis may be affected by the parthanatos process. Endocytosis is a mechanism for cellular uptake of membrane-impermeant nutrients. Rab5, a small G-protein, is associated with the plasma membrane and early endosomes. Once activated by binding GTP, Rab5 recruits its effectors to early endosomes and regulates their fusion. Here, we report that after DNA damage, PARP1-generated PAR binds to Rab5, suppressing its activity. As a result, Rab5 is dissociated from endosomal vesicles, inhibiting the uptake of membrane-impermeant nutrients. This PARP1-dependent inhibition of nutrient uptake leads to cell starvation and death. It thus appears that this mechanism may represent a novel parthanatos pathway.


Introduction
Parthanatos is a poly(ADP-ribose) polymerase1 (PARP1)-dependent programmed cell death, which occurs after DNA damage [1][2][3]. PARP1 contains zinc-finger domains that recognize DNA strand breaks and initiate poly(ADP-ribosyl)ation of nuclear proteins, including histones and PARP1 itself [4,5]. When the damage is mild, this poly(ADP-ribosyl)ation of nuclear proteins contributes to DNA repair by recruiting the repair machinery, and is rapidly terminated by the action of nuclear PAR-degrading enzymes such as poly(ADPribose) glycohydrolase (PARG) and ADP-ribosyl-acceptor hydrolase 3 (ARH3) [6,7]. On the other hand, when DNA damage is severe, some of the excess poly(ADP-ribose) (PAR) synthesized by overactivated PARP1 is cleaved by PARG, generating protein-free PAR polymer, which is subsequently translocated to the cytoplasm [8][9][10]. In the cytoplasm, PAR binds non-covalently to cytoplasmic and mitochondrial proteins, including apoptosis-inducing factor (AIF), leading to its release from the mitochondrial membrane [3,10,11]. Because AIF possesses a nuclear localization sequence, it is translocated to the nucleus where it interacts with DNase, resulting in large-scale DNA fragmentation and cell death [12]. In addition, PAR also interacts with hexokinase, a key glycolytic enzyme [13,14]. Binding to PAR
The inhibition of OVA uptake developed at least 20 min after exposure to MNNG in NIH3T3 cells (Figure 2A,B). Simultaneous addition of MNNG with OVA-FITC did not affect OVA uptake ( Figure 2C,D), suggesting PARP1-mediated inhibition of OVA uptake requires minutes to develop. Once a substance is taken up by endocytosis, a portion of it is returned to the extracellular environment via endosomal recycling. To determine whether PARP1 activation accelerates endosomal recycling, OVA-FITC was added to NIH3T3 cells for 1 h before addition of MNNG. Exposure to MNNG after OVA-FITC uptake did not affect the proportion of cells that contained OVA-FITC ( Figure 2E), indicating that PARP1 inhibits OVA uptake rather than promoting the recycling pathway. Uptake of OVA as well as transferrin (Tf) is mainly via receptor-mediated endocytosis and macropinocytosis, while uptake of smaller molecules (e.g., dextran) is accomplished by macropinocytosis [27]. PARP1 also inhibited uptake of both Tf-Alexa647 and dextran-rhodamine ( Figure 2F,G), indicating that PARP1 activation following exposure to MNNG inhibits receptor-mediated endocytosis and macropinocytosis. The inhibition of OVA uptake developed at least 20 min after exposure to MNNG in NIH3T3 cells (Figure 2A,B). Simultaneous addition of MNNG with OVA-FITC did not affect OVA uptake ( Figure 2C,D), suggesting PARP1-mediated inhibition of OVA uptake requires minutes to develop. Once a substance is taken up by endocytosis, a portion of it is returned to the extracellular environment via endosomal recycling. To determine whether PARP1 activation accelerates endosomal recycling, OVA-FITC was added to NIH3T3 cells for 1 h before addition of MNNG. Exposure to MNNG after OVA-FITC uptake did not affect the proportion of cells that contained OVA-FITC ( Figure 2E), indicating that PARP1 inhibits OVA uptake rather than promoting the recycling pathway. Uptake of OVA as well as transferrin (Tf) is mainly via receptor-mediated endocytosis and macropinocytosis, while uptake of smaller molecules (e.g., dextran) is accomplished by macropinocytosis [27]. PARP1 also inhibited uptake of both Tf-Alexa647 and dextran-rhodamine ( Figure 2F,G), indicating that PARP1 activation following exposure to MNNG inhibits receptor-mediated endocytosis and macropinocytosis. ). NIH3T3 cells were pretreated with or without PJ34 (10 μM) for 10 min, then exposed to MNNG (500 μM, 20 min), after which they were incubated with OVA-FITC (50 μg/mL) then chased NIH3T3 cells were pretreated with or without PJ34 (10 µM) for 10 min, then exposed to MNNG (500 µM, 20 min), after which they were incubated with OVA-FITC (50 µg/mL) then chased with OVA (100 µg/mL) for the indicated times. Shown are means ± SEM (n = 3). p < 0.05 at >20 min (left) and >60 min (right). (B) Subcellular localization of OVA-Alexa647. NIH3T3 cells were pretreated with or without PJ34 (10 µM) for 10 min and then exposed to MNNG (500 µM, 20 min) before incubation with OVA-Alexa647 (50 µg/mL) and chased with OVA (100 µg/mL) for the indicated times. Scale bar: 10 µm. (C) OVA-FITC uptake following simultaneous exposure to MNNG and OVA-FITC. NIH3T3 cells were pretreated with or without PJ34 (10 µM) for 10 min, then exposed to MNNG (500 µM) and OVA-FITC (50 µg/mL, green) for 20 min. Nuclei were stained with DAPI (blue). Scale bar: 10 µm. (D) Effect of simultaneous addition of MNNG and OVA-FITC on OVA-FITC uptake. With the same procedure as in panel (C), OVA-FITC + cells (left) and MFI of OVA-FITC (right) were counted using flow cytometry. Shown are means ± SEM (n = 3). (E) Effect of endosomal recycling on OVA-FITC uptake (the percentages of OVA-FITC + cells (upper) and MFI of OVA-FITC in NIH3T3 cells (lower)). After incubation with OVA-FITC (50 µg/mL, 1 h), NIH3T3 cells were exposed to MNNG (500 µM, 20 min) with OVA (100 µg/mL, 1 h). Shown are means ± SEM (n = 3). (F) Tf-Alexa647 uptake (the percentages of Tf-Alexa647 + cells (upper) and MFI of Tf-Alexa647 in NIH3T3 cells (lower)). NIH3T3 cells were exposed to MNNG (500 µM, 20 min) before incubation with Tf-Alexa647 (15 µg/mL, 1 h). Shown are means ± SEM (n = 3).** p < 0.01, *** p < 0.001 (G) Dextran-rhodamine uptake (the percentages of Dextran-rhodamine + cells (upper) and MFI of Dextran-rhodamine in NIH3T3 cells (lower)). NIH3T3 cells were exposed to MNNG (500 µM, 20 min) before incubation with dextran-rhodamine (50 µg/mL, 1 h). Shown are means ± SEM (n = 3). *** p < 0.001 (H) Shortand long-term cytotoxic effects of MNNG. NIH3T3 cells were pretreated for 10 min with or without PJ34 (10 µM) before cell viability was measured after 1-h (black) or 24-h (red) exposure to MNNG at the indicated concentrations. Shown are means ± SEM (n = 3). p < 0.001 between 1-h and 24-h at concentrations > 100 µM in control group. (I) MNNG-induced cytotoxicity during starvation. NIH3T3 cells were pretreated for 10 min with or without PJ34 (10 µM), then exposed to MNNG (20 min) at the indicated concentrations before incubation for 24 h in serum-containing (black) or serum-free (red) medium. Shown are means ± SEM (n = 3). p < 0.001 between control and starvation at concentrations >100 µM in the control group. Because MNNG-induced PARP1 activation triggers parthanatos, we assessed the effect of PARP1-mediated inhibition of OVA uptake on cell viability. NIH3T3 cells exposed to MNNG for 24 h exhibited diminished viability, and this effect was suppressed by pretreatment with the PARP1 inhibitor PJ34 ( Figure 2H). In contrast, exposure to MNNG for 1 h was rarely cytotoxic ( Figure 2H). This suggests PARP1-mediated inhibition of OVA uptake occurs prior to cell death, as parthanatos only occurs after long-term exposure to MNNG.
Endocytosis contributes to cell survival by supplying cells with nutrients and macromolecules that cannot permeate the cell membranes and by controlling receptor signaling to adapt cellular stress [20,21]. When NIH3T3 cells were starved in serum-free Dulbecco's modified Eagle's medium (DMEM), exposure to MNNG exacerbated the decline in cell viability as compared to cells in DMEM with 10% fetal bovine serum (FBS) ( Figure 2I). Thus, PARP1-mediated inhibition of endocytosis results in cell death, which may be a novel pathway in PARP1-dependent parthanatos.
To identify the molecular mechanism underlying PARP1-mediated inhibition of OVA uptake, we observed the subcellular localization of endocytosis-related proteins ( Figure S2A). Exposure to MNNG altered the distribution of the early endosomal proteins (Rab5, Rabex5, Rabenosyn5 and EEA1), but not late-endosomal (Rab7) or lysosomal proteins (Lamp1 and 2 and Cathepsin S), from a punctate pattern to uniform distribution in the cytoplasm ( Figures 3A and S2B). Stable expression of PARP1 shRNA, or pretreatment with PJ34, suppressed the MNNG-induced changes in the localization of early endosomal proteins ( Figures 3A and S2B). The distribution of Flag-tagged Vps34, a class III phosphatidylinositol (PI)-3 kinase, was also altered upon PARP1 activation ( Figures 3B and S2C). Time-lapse imaging of red fluorescent protein (RFP)-EEA1 and blue fluorescent protein (BFP)-Rab5 af-ter DNA damage induced by MNNG ( Figure S3A,B) or laser micro-irradiation of the nuclei ( Figure S3C,D) revealed their similar redistribution, which was suppressed by pretreatment with PJ34.  The changes in the subcellular localization of Rab5 and EEA1 appeared 20 min after exposure to MNNG, which was consistent with the time required for PAR synthesis in the nucleus and its translocation to the cytoplasm ( Figure 3C). Moreover, after 10-min exposure to MNNG, some punctate PAR signals in the cytoplasm were colocalized with Rab5 and EEA1 ( Figure 3D). Western blot analysis revealed that exposure to MNNG did not induce degradation of these endosomal proteins ( Figure 3E). Instead, it enhanced their dissociation from EEs. The changes in the subcellular localization of Rab5 and EEA1 appeared 20 min after exposure to MNNG, which was consistent with the time required for PAR synthesis in the nucleus and its translocation to the cytoplasm ( Figure 3C). Moreover, after 10-min exposure to MNNG, some punctate PAR signals in the cytoplasm were colocalized with Rab5 and EEA1 ( Figure 3D). Western blot analysis revealed that exposure to MNNG did not induce degradation of these endosomal proteins ( Figure 3E). Instead, it enhanced their dissociation from EEs.

PAR Binds to the PBM of Rab5 in Cells Exposed to MNNG
Because only proteins localized in EEs were dissociated upon PARP1 activation, we investigated whether Rab5, a master regulator of early endosomal fusion, is inactivated through PAR binding. To detect PAR-Rab5 binding, pull-down assays using GST-Af1521 macrodomain were performed to isolate PAR-bound proteins ( Figure S4A) [28]. Assays using GST-Af1521 macrodomain, but not GST, revealed PAR binding to RFP-Rab5 in HeLa cells exposed to MNNG for 10 min (Figures 4A,B and S4B). The interaction was blocked, when PAR synthesis was suppressed by expression of PARP1 shRNA or pretreatment with PJ34 ( Figures 4A,B and S4B-D). To test whether PAR binding blocks Rab5 activity, the levels of GTP-bound Rab5 were measured in pull-down assays using GST-Rab5-binding domain (R5BD) from Rabaptin5 ( Figure S4E) [29]. The assays revealed that levels of active RFP-Rab5 were diminished in cells exposed to MNNG for 10 min (Figures 4C,D and S4F-H). By contrast, RFP-Rab5Q79A, a dominant-active Rab5 mutant, remained in the GTP-bound form in those cells ( Figure 4E,F). Exposure to MNNG also did not affect the subcellular localization of RFP-Rab5Q79A ( Figure 4G). Moreover, expression of RFP-Rab5Q79A slightly reduced the amount of OVA-FITC incorporated into cells following exposure to MNNG, but did not alter the percentage of cells that took up OVA-FITC ( Figure 4H). These results indicate that PARP1-generated PAR binds Rab5 in the cytoplasm, thereby suppressing GDP-to-GTP exchange and resulting in loss of Rab5 activity.
GTP-bound Rab5 recruits Vps34 to EEs where it catalyzes the phosphorylation of PI to phosphatidylinositol-3-phosphate (PI3P) [30]. PI3P in endosomal vesicles recruits EEA1 and rabenosyn5, which contains an FYVE domain [31]. Because PARP1 activation also leads to dissociation of Vps34 from EEs along with Rab5 ( Figure 3B), we tested whether phosphatidylinositol phosphate (PIP) levels in membrane fractions were reduced by MNNG. Electrospray ionization mass spectrometry revealed that exposing cells to MNNG for 20 min led to PARP1-dependent decreases in the PIP-to-PI ratio ( Figures 4I,J and S4I). Consistent with that finding, MNNG induced dissociation of green fluorescent protein (GFP)-FYVE, a PI3P marker, from EEs ( Figures 4K and S4J), suggesting that PARP1 activation reduces PI3P levels in EEs.
were diminished in cells exposed to MNNG for 10 min (Figures 4C,D and S4F-H). By contrast, RFP-Rab5Q79A, a dominant-active Rab5 mutant, remained in the GTP-bound form in those cells ( Figure 4E,F). Exposure to MNNG also did not affect the subcellular localization of RFP-Rab5Q79A ( Figure 4G). Moreover, expression of RFP-Rab5Q79A slightly reduced the amount of OVA-FITC incorporated into cells following exposure to MNNG, but did not alter the percentage of cells that took up OVA-FITC ( Figure 4H). These results indicate that PARP1-generated PAR binds Rab5 in the cytoplasm, thereby suppressing GDP-to-GTP exchange and resulting in loss of Rab5 activity.

A putative PAR-binding motif ([HKR] 1 -X 2 -X 3 -[AIQVY] 4 -[KR] 5 -[KR] 6 -[AILV] 7 -[FILPV] 8 )
has been identified through proteomic analysis [25]. Notably, the positively charged amino acids of the 5th and 6th arginine and/or lysine residues ([KR] 5 -[KR] 6 ) are important for binding to the negatively charged phosphate group of PAR. Rab5 appears to have two putative PAR-binding motifs located at around amino acid residues 140 and 180, respectively ( Figure 5A). To identify the PAR-binding site on Rab5, RFP-Rab5 K140AK141A (Rab5 140 mutant) and K180AR181A (Rab5 180 mutant) mutants were prepared. Like RFP-Rab5 wild-type (WT), both the Rab5 140 and 180 mutants showed a punctate distribution in the cytoplasm under unstimulated conditions ( Figure 5B). However, MNNG induced dissociation of Rab5 WT and the Rab5 140 mutant from EEs, whereas the Rab5 180 mutant remained in EEs ( Figure 5B). In contrast to RFP-Rab5 WT and the Rab5 140 mutant, the RFP-Rab5 180 mutant did not bind PAR, and levels of the GTP-bound form were unaffected by MNNG ( Figure 5C-F). In addition, expression of the RFP-Rab5 180 mutant restored OVA-Alexa647 uptake in cells exposed to MNNG ( Figure 5G,H). These results indicate that PAR, generated by PARP1 in response to DNA damage, and then translocated to the cytoplasm, binds the PAR-binding motif near amino acid residue 180 of Rab5, which results in loss of its activity. These results indicate that PAR, generated by PARP1 in response to DNA damage, and then translocated to the cytoplasm, binds the PAR-binding motif near amino acid residue 180 of Rab5, which results in loss of its activity.  Ratios of PAR-bound Rab5 (pull-down) to Rab5 (input) were normalized to control. Shown are means ± SEM (n = 3). *** p < 0.001 vs. WT. (E) Rab5 activity assay. After exposure to MNNG (100 µM, 20 min), HeLa cells expressing RFP-Rab5 WT or its mutants were subjected to first pull-down assays using R5BD-GST and then to Western blotting using the indicated antibodies. (F) Relative Rab5 activity. Ratios of GTP-bound Rab5 (pull-down) to Rab5 (input) were normalized to control. Shown are means ± SEM (n = 3). ** p < 0.01. (G) Histogram of OVA-Alexa647 + cells. HeLa cells expressing RFP-Rab5 WT or its mutants were exposed to MNNG (100 µM, 20 min) before incubation with OVA-FITC (50 µg/mL, 1 h) (green) and then chased with OVA (100 µg/mL, 1 h). OVA-Alexa647 + cells (within magenta region) were counted using flow cytometry.

Rab5 and Its Effectors Are Dissociated from EEs in Neuronal Cells after PARP1 Activation
Parthanatos is frequently the cause of neuronal cell death [1,8]. We therefore tested whether PARP1-dependent inhibition of OVA uptake occurs in primary cultures of hippocampal neurons. Exposure to MNNG resulted in dissociation of Rab5 and EEA1 from EEs within the neurites and soma of neurons ( Figure 6A) and inhibition of OVA uptake ( Figure 6B). These effects were blocked by pretreatment with PJ34. Thus, under conditions where parthanatos occurs in neurons, PARP1 inhibits nutrient uptake by suppressing EE maturation. EEs within the neurites and soma of neurons ( Figure 6A) and inhibition of OVA uptake ( Figure 6B). These effects were blocked by pretreatment with PJ34. Thus, under conditions where parthanatos occurs in neurons, PARP1 inhibits nutrient uptake by suppressing EE maturation. were pretreated for 10 min with or without PJ34 (10 μM), then exposed to MNNG (500 μM, 20 min) before staining with anti-MAP2 antibodies (red) and anti-Rab5 (green, left) or anti-EEA1 (green, right). (B) OVA-Alexa uptake by primary hippocampal neurons. Neurons were pretreated for 10 min with or without PJ34 (10 μM), then exposed to MNNG (500 μM, 20 min) before addition to OVA-Alexa647 (50 μg/mL, 1 h). Neurites and nuclei were marked with MAP2 (green) and DAPI (blue), respectively. Scale bar: 10 μm.

Discussion
In this study, we found that Rab5 and its effectors dissociate from EEs upon PARP1 activation. Thus, PARP1-dependent suppression of OVA, Tf and dextran uptake after DNA damage apparently reflects the suppression of EE maturation and fusion that results from dissociation of Rab5 and its effectors. This finding is consistent with the fact that, in vivo, in adult mouse liver Rab5 knockdown using siRNA markedly reduces numbers of EEs, late endosomes and lysosomes and is associated with suppression of endocytosis of low-density lipoprotein [33]. After insertion into endosomal membranes, GTP-bound Rab5 binds to Rabaptin5, which recruits Rabex5, a Rab5 GEF, to maintain the GTP-bound state of Rab5 [34,35]. PARP1 activation reduces the GTP-bound state in cells expressing RFP-Rab5 WT but not a constitutively active form of Rab5. We therefore speculate that in the event of DNA damage, PAR produced by PARP1 binds to the PAR-binding motif in Figure 6. PARP1 inhibits endocytosis by dissociating Rab5 and EEA1 in hippocampal primary neurons. (A) Subcellular localization of Rab5 an EEA1 within neurites. Primary hippocampal neurons were pretreated for 10 min with or without PJ34 (10 µM), then exposed to MNNG (500 µM, 20 min) before staining with anti-MAP2 antibodies (red) and anti-Rab5 (green, left) or anti-EEA1 (green, right). (B) OVA-Alexa uptake by primary hippocampal neurons. Neurons were pretreated for 10 min with or without PJ34 (10 µM), then exposed to MNNG (500 µM, 20 min) before addition to OVA-Alexa647 (50 µg/mL, 1 h). Neurites and nuclei were marked with MAP2 (green) and DAPI (blue), respectively. Scale bar: 10 µm.

Discussion
In this study, we found that Rab5 and its effectors dissociate from EEs upon PARP1 activation. Thus, PARP1-dependent suppression of OVA, Tf and dextran uptake after DNA damage apparently reflects the suppression of EE maturation and fusion that results from dissociation of Rab5 and its effectors. This finding is consistent with the fact that, in vivo, in adult mouse liver Rab5 knockdown using siRNA markedly reduces numbers of EEs, late endosomes and lysosomes and is associated with suppression of endocytosis of low-density lipoprotein [33]. After insertion into endosomal membranes, GTP-bound Rab5 binds to Rabaptin5, which recruits Rabex5, a Rab5 GEF, to maintain the GTP-bound state of Rab5 [34,35]. PARP1 activation reduces the GTP-bound state in cells expressing RFP-Rab5 WT but not a constitutively active form of Rab5. We therefore speculate that in the event of DNA damage, PAR produced by PARP1 binds to the PAR-binding motif in its C-terminal region of Rab5, preventing Rab5 from maintaining its GTP-bound state. Rab5 consists of six central β-sheets (β1-6) surrounded by five α-helices (α1-α5) [36,37]. The PAR-binding motif of Rab5 is located within α5-helix, on the surface of the protein, opposite the β-sheet required for the binding Rabaptin5 [34]. The binding of PAR to Rab5 induces a conformational change in the protein that inhibits its binding to Rabaptin5. In fact, because the Rab5-binding domain used to detect GTP-bound Rab5 in this study is derived from Rabaptin5, the results of the GST pull-down assays with R5BD are consistent with the dissociation of the Rab5-Rabaptin5 interaction upon PAR binding. Alternatively, the PAR-binding motif of Rab5 is in close proximity to the two prenylated cysteine residues in the C-terminal region [23], so that PAR binding may also inhibit their modification. As a result, PAR binding may prevent Rab5 from translocating to the EE membrane and inhibit its GEF-mediated GTP exchange.
Vps34 binds to GTP-bound Rab5 and catalyzes the phosphorylation of PI to PI3P in EEs [30]. EEA1 binds to GTP-bound Rab5 and to PI3P in the EE membrane via the FYVE domain, thereby associating EEs with each other [31]. In addition, GTP-bound Rab5 fuses EEs together by accumulating Rabenosyn-5-Vps45 complexes, which modulate the accumulation of vesicle fusion-regulating SNAREs [38]. The dissociation of activated Rab5 from EEs resulted in the dissociation of Vps34 and a decrease in the PI3P content of EEs. This suggests that dissociation of EEA1 and Rabenosyn5 from vesicles follows from the dissociation of Rab5 from EEs.
GDI binds to GDP-bound Rab5 and prevents it from localizing at the membrane. The activity of GDI is regulated by p38 MAPK [32]. The fact that p38 inhibition did not suppress the inhibition of OVA uptake following PARP1 activation, and that GDI localization was unchanged in cells exposed to MNNG, indicates that GDI does not participate in the inactivation of Rab5 or its dissociation from EEs following PARP1 activation.
Endocytosis contributes to cell survival by mediating the uptake of proteins and other membrane-impermeant nutrients [21]. In addition, it controls receptor activation and signaling [39,40]. Endocytosis reduces the number of receptors available for extracellular ligands, which attenuates signaling triggered by the plasma membrane. On the other hand, many types of receptors require endocytosis to interact with downstream effectors for sustained activation of signaling. In this study, we found that PARP1-dependent inhibition of Rab5 activity promotes cell death under starvation conditions. This may result from inhibition of nutrient uptake and receptor signaling involved in cell survival and proliferation, which may be a novel parthanatos pathway. Parthanatos has been implicated in Parkinson's disease, Alzheimer's disease, and neuronal death during cerebral ischemiareperfusion [15][16][17][18][19]. Although the role of PARP1-dependent inhibition of nutrient uptake remains unclear, our findings suggest its potential involvement in the induction of neuronal cell death during these pathological conditions.

Cell Culture
NIH3T3 and HeLa cells were incubated in DMEM containing 10% FBS, 100 units of penicillin, and 100 µg/mL streptomycin at 37 • C in a humidified atmosphere with 5% CO 2 . shRNA plasmids targeting human or mouse PARP1 or scrambled shRNA (Origene) were introduced into NIH3T3 and HeLa cells using Lipofectamine 3000 transfection reagent according to manufacturer's instructions. Cells stably expressing shRNA plasmids were selected in medium containing 1 µg/mL puromycin.
Primary hippocampal neurons were isolated using neuron dissociation solution (Sumiron, Osaka, Japan) and cultured in neurobasal medium with B-27 supplement on polyd-lysine-coated dishes. Half the volume of the medium was replaced every 5 days. The protocols used in this study were approved by the Ethical Committee of Doshisha Women's College of Liberal Arts (Nos. Y15-027, Y16-30, Y17-031).

Cell Viability Assays
NIH3T3 and HeLa cells (1 × 10 4 cells) seeded onto 96-well plates were incubated for 30 min with a PARP inhibitor and then exposed to MNNG. Cell numbers were then counted using cell counting reagent SF (Nacalai Tesque) according to the manufacturer's instructions by measuring the absorbance at 450 nm (SpectraMax M5 Microplate Reader, Molecular Devices, San Jose, CA, USA).

Live-Cell Imaging after DNA Damage
NIH3T3 cells (1 × 10 5 cells) seeded onto glass-bottomed dished were transfected using Lipofectamine 3000 according to the manufacturer's instructions. After incubation for 1 day, fluorescence was observed with a confocal microscope (Zeiss LSM 700 Meta) equipped with an oil-immersion objective (63×, numerical aperture = 1.4). Hoechst 33258 (10 µg/mL; Dojindo, Kumamoto, Japan) was added to enhance DNA damage. Micro-irradiation of whole nuclei was carried out with a 405 nm diode laser set to 100% transmission [46]. The microscope was equipped with a heated environmental chamber set to 37 • C. Images were taken every 30 s for 40 min.

Western Blotting
Cells (3 × 10 5 cells) seeded onto 6-well plates were incubated (1 day, 37 • C) in DMEM with 10% FBS. Cell lysates were prepared with 2% SDS in 20 mM Tris-HCl (pH 7.4) containing complete protease inhibitor cocktail (Roche, Basel, Switzerland). After adjustment of the protein concentration using a BCA kit (Thermo Fisher Scientific), cell lysates were subjected to Bis-Tris SDS-PAGE (Thermo Fisher Scientific) and then transferred to nitrocellulose membranes (Thermo Fisher Scientific). The membranes were blocked with Blocking One for 30 min at room temperature and then incubated with primary antibodies. After incubation with HRP-conjugated anti-mouse or rabbit IgG secondary antibodies, an ECL system (Amersham Imager 600, Cytive, Marlborough, MA, USA) was used for detection.

Extraction of the Membrane Fraction and Lipid Isolation and Measurement of PI and PIP with Mass Spectrometry
Cells were precipitated with 0.5 M trichloroacetic acid (TCA). After washing with 5% TCA with 1 mM EDTA, membrane fractions were lysed with MeOH:CHCl 3 :12N HCl (80:40:1) and then separated by addition of CHCl3:0.1 N HCl (1:1). The organic layer was collected and concentrated using a vacuum evaporator. 9-Aminoacridine (Sigma-Aldrich, St. Louis, MI, USA) was used as a matrix. Mass spectrometry was performed using matrix-assisted laser desorption and ionization/time of flight mass spectrometry (MALDI-TOF-MS) (ultrafleXtreme, Bruker, Billerica, MA, USA) [47]. The mass spectra of phosphatidylinositol (PI) and phosphatidylinositol phosphate (PIP) were acquired in the negative ion reflector mode.

Statistical Analysis
Statistical analysis was performed using Sigmaplot 13 (Systat Software Inc., San Jose, CA, USA). Significance was determined using Student's t-test between two samples or one-way and two-way ANOVA with post hoc Tukey's test for three or more groups. Data are means ± S.E.M of values from the indicated number of experiments. p values < 0.05 were considered significant. All representative experiments were repeated three times.