Arabidopsis PARG1 is the key factor promoting cell survival among the enzymes regulating post-translational poly(ADP-ribosyl)ation

Poly(ADP-ribosyl)ation is a reversible post-translational modification of proteins, characterized by the addition of poly(ADP-ribose) (PAR) to proteins by poly(ADP-ribose) polymerase (PARP), and removal of PAR by poly(ADP-ribose) glycohydrolase (PARG). Three PARPs and two PARGs have been found in Arabidopsis, but their respective roles are not fully understood. In this study, the functions of each PARP and PARG in DNA repair were analyzed based on their mutant phenotypes under genotoxic stresses. Double or triple mutant analysis revealed that PARP1 and PARP2, but not PARP3, play a similar but not critical role in DNA repair in Arabidopsis seedlings. PARG1 and PARG2 play an essential and a minor role, respectively under the same conditions. Mutation of PARG1 results in increased DNA damage level and enhanced cell death in plants after bleomycin treatment. PARG1 expression is induced primarily in root and shoot meristems by bleomycin and induction of PARG1 is dependent on ATM and ATR kinases. PARG1 also antagonistically modulates the DNA repair process by preventing the over-induction of DNA repair genes. Our study determined the contribution of each PARP and PARG member in DNA repair and indicated that PARG1 plays a critical role in this process.

Scientific RepoRts | 5:15892 | DOi: 10.1038/srep15892 We grew these mutants on the same plate to compare their phenotypes. Two kinds of genotoxic agents, bleomycin and methyl methane sulfonate (MMS), were used in the assays, respectively. Bleomycin is a glycopeptide which mainly induces DSBs 35 and MMS is a monofunctional alkylating agent that induces N-alkyl lesions and SSBs 36 . DSBs and SSBs can activate different DNA repair pathways 37,38 . DSBs are the most severe damage type since chromosomal breakage can lead to cell death 32,38 . On control plates, all mutants grew normal and were indistinguishable from each other (Fig. 1A). On the plate with 50 μ g/ml bleomycin, parg1-2 and parg1-4 turned almost completely yellow (Fig. 1B,D), while parp1, parp2 and parg2-1 were only mildly sensitive to bleomycin compared with Col-0, and parg2-2 and parp3 responded similarly as the wild type control (Fig. 1B). MMS blocked the growth of parg1-2 and parg1-4, but did not induce yellowing (Fig. 1C), so we measured the fresh weight of each mutant and compared them with that of the Col-0 (Fig. 1E). Most of the single mutants exhibited slight sensitivity to MMS except parp3, which had no obvious phenotype, and parg1-4, which was hypersensitive to MMS. Surprisingly, another mutant allele of PARG1, parg1-2, is only slightly sensitive to MMS (Fig. 1C,E).
Taken together, the results revealed that among all mutants, the parg1-4 mutant displayed the most severe phenotype while parp3 had no obvious phenotype compared to the wild type control when challenged by DNA SSB or DSB inducing agents.

Phenotypic analysis revealed the relationships between PARP and PARG family members.
It has been reported that among the three PARPs, when PARP1 or PARP2 is absent, the expression levels of the other two PARPs are up-regulated 12 . We also found that when PARG1 is knocked out, the remaining PARG2 gene is up-regulated, and vice versa ( Supplementary Fig. S3A). Therefore, functional redundancy might exist between the same family members. To further analyze it, we constructed parp-1parp2 (p1p2) double and parp1parp2parp3 (p1p2p3) triple mutants. We were unable to generate double mutant of the closely linked PARG1 and PARG2 genes, so RNA interference was used to inhibit the expression of PARG2 in the parg1-4 mutant background (g1g2). All these mutants grew normally on 1/2 MS plate ( Fig. 2A), while phenotypic comparison by bleomycin and MMS treatments revealed that the p1p2p3 triple mutant was not more sensitive than the p1p2 double mutant (Fig. 2B-E), indicating that either PARP3 does not perform the same role as PARP1 and PARP2, or is not an active enzyme. This speculation is also supported by the analysis result of the parp3 single mutant, which was not sensitive to bleomycin and MMS. Two g1g2 lines, g1g2-8 and g1g2-9, were more sensitive to bleomycin and All experiments were done for at least three times and similar results were obtained. The data were presented as means of three replicates ± SE. Significant differences (t-test) compared to Col-0 under the same conditions are indicated by asterisks: *P < 0.05; **P < 0.01. MMS than wild type control (Fig. 2B,C). Inhibition of PARG2 expression in the parg1-4 mutant did not enhance the sensitivity of parg1-4 to bleomycin, which was already severe, but did enhance its resistance to MMS (Fig. 2B,C).
The phenotype of the g1g2 mutant was much more severe than that of the p1p2 and p1p2p3 mutants on both bleomycin and MMS plates (Fig. 2B,C), indicating a more detrimental effect of PARG mutation to DNA repair than that of PARPs mutation. Between the two PARGs, PARG1 is the major one determining plant growth and survival under bleomycin and MMS treatments. We therefore focused on the study of PARG1 and mainly used parg1-4 mutant for the following experiments due to its consistently strong phenotype on bleomycin and MMS plates. When a genomic fragment spanning from 1379 bp upstream of the translational start site to 969 bp downstream of stop codon of PARG1 was introduced into parg1-4, the bleomycin-and MMS-sensitivity phenotype of parg1-4 were rescued ( Supplementary  Fig. S4), demonstrating that loss of function of PARG1 caused the phenotypes. To simplify the analysis, we used bleomycin to induce DNA damages in the following experiments.

PARG1 degrades poly(ADP-ribose) in vitro.
PARG is considered as the downstream enzyme of PARP, but disruption of PARG1 caused more severe phenotype than that of p1p2 double and p1p2p3 triple mutant under genotoxic stress. It raised the question whether the phenotype is directly caused by the loss of PARG activity. To answer this question, we first examined the poly(ADP-ribose) glycohydrolase activity of PARG1 in vitro using recombinant PARG1 protein. We cloned the cDNA of PARG1 and expressed PARG1 as a glutathione-S-transferase (GST) fusion protein in E. coli. The purified GST-PARG1 was used for the activity assay.
Arabidopsis PARP1 can be self-modified by adding varying numbers of ADP-riboses onto itself in the presence of broken DNA and NAD +14 . The auto-PARylated protein was detected as an upwardly shifted smear on SDS-PAGE gel (Fig. 3A upper and middle panel). This auto-modified PARP1 was used as a PARylated protein substrate for the PARG1 activity assay. PAR was detected by anti-PAR antibody ( Fig. 3A bottom panel). This antibody tends to recognize big ADP-ribose polymers due to their stronger immunogenicity. When the auto-modified PARP1 was incubated with the recombinant GST-PARG1 protein, the upward smear of the PARP1 disappeared and the band became concentrated again at the original size of PARP1 on the SDS-PAGE gel within one minute of adding GST-PARG1 ( Fig. 3B middle panel), and PAR signal on the protein was concomitantly reduced (Fig. 3B bottom panel), indicating that PARG1 can rapidly degrade the poly(ADP-ribose) on the modified PARP1 and restore the size of the protein (Fig. 3B middle panel). For comparison, a mutant protein E273N (Glu273 → Gln) with a mutation in the conserved GGG-X6-9-QEE motif of PARG1 was generated and subjected to the same All experiments were done for at least three times and similar results were obtained. The data were presented as means of three replicates ± SE. Significant differences (t-test) compared to Col-0 under the same conditions are indicated by asterisks: *P < 0.05; **P < 0.01. assay. The mutation completely destroyed the PAR hydrolase activity (Fig. 3B). As a control, the GST tag did not cause any mobility change of the PARylated PARP1 (Fig. 3C). When the commercial PAR was incubated with the recombinant PARG1 protein for different time periods and then detected with anti-PAR antibody, the PAR signal also disappeared gradually (Fig. 3D). These results demonstrated that PARG1 has a PAR hydrolase activity not only against PARylated proteins, but also against free PAR.
Loss of PARG1 induces cell death under genotoxic stress. When non-lethal dose of bleomycin was included in the medium, the parg1-4 mutant grew more slowly and had fewer true leaves than wild type seedlings, reflected by a lower fresh weight of parg1 ( Supplementary Fig. S5A,B). Leaf color of the parg1 mutant turned yellow gradually with the increase of bleomycin concentration ( Supplementary Fig.  S5C), suggesting that chlorophyll was degraded and cell death might happen. Trypan blue was used to stain the tissues for detection of cell death 39 . The staining results indicated that bleomycin induced more cell death in the leaves of parg1 than those of wild type (Fig. 4A). Furthermore, stronger H 2 O 2 accumulation could be detected in the mutant cotyledons (Fig. 4B) when stained by 3, 3'-diaminobenzidine (DAB), which yields a brown precipitate in the presence of H 2 O 2 40 . Thus, the genotoxin treatment also led to a more oxidized status in the mutant than in wild type.
In addition, the root growth of parg1-4 was also investigated. Under normal conditions, the mutant had no significant difference from Col-0 (Fig. 5A). However, bleomycin treatment caused more cell death in the root meristem of the mutant than that of wild type, shown by propidium iodide (PI) staining (Fig. 5B). The cell death led to significant reduction of the primary root growth of the mutant compared to that of wild type (Fig. 5C). When the time period of the genotoxic stress was extended, a much shorter meristematic zone was observed at the root tip of the parg1-4 mutant (Fig. 5D), which might be due to the enhanced cell death in the parg1-4 root meristem. Meanwhile, the epidermal cells of the parg1-4 mutant were also more deformed and enlarged than that of wild type (Fig. 5D).
These results indicated that disruption of PARG1 causes more cell death in both leaves and roots when the plants were treated by bleomycin.

PARG1 regulates poly(ADP-ribose) level in vivo.
To test whether the phenotype of the parg1-4 mutant was caused by over-accumulation of poly(ADP-ribose), we added 3-AB, a cell permeable PARP inhibitor 29 into the media to repress PARP activity. 3-AB reversed the phenotype of parg1-4 and restored the seedling growth of the parg1-4 mutant close to that of wild type under bleomycin treatment (Fig. 6A). In addition, the death phenotype of parg1-4 seedlings under high level of bleomycin was also rescued by 3-AB ( Supplementary Fig. S6). These results suggested that PARG1 hydrolyzes PAR synthesized by PARPs in vivo and excessive PAR in the parg1-4 mutant is the reason for growth arrest and cell death under bleomycin treatment. However, parg1-4 was a little smaller than Col-0 on plates containing bleomycin and 3-AB (Fig. 6A, Supplementary Fig. S6). This may be due to incomplete suppression of PARP activity by 3-AB in vivo, or that PARG1 has other functions not affected by 3-AB.
To further confirm the biochemical activity of PARG1 in vivo, we examined the PAR levels in parg1-4 and wild-type plants by western blot analysis using anti-PAR antibody. The results revealed that the levels of PAR were relatively low in both wild type and parg1-4 plants grown under normal conditions (Fig. 6C). When treated with bleomycin, the parg1-4 mutant accumulated more PAR than wild type, and 3-AB reduced PAR in both wild type and the parg1-4 mutant (Fig. 6C). These results once again indicated that PARG1 regulates PAR level in vivo.

PARG1 is induced primarily in mitotically active cells. Quantitative RT-PCR was performed to
investigate the expression level changes of PARG1 under genotoxic stress. The transcript level of PARG1 in the seedlings subjected to a brief bleomycin treatment was examined ( Supplementary Fig. S7). The results showed that PARG1 expression was induced by bleomycin. GUS staining of pPARG1::GUS transgenic line showed that PARG1 expression was initially induced in the root and shoot meristems ( Supplementary Fig. S7B), then extended to other tissues.
Loss-of-PARG1 leads to the transcriptional up-regulation of DNA repair genes and increase of cellular DNA damage level. Eukaryotic organisms have two major pathways for repairing DNA double strand breaks, the homologous recombination (HR) and the non-homologous end joining (NHEJ) pathways 41,42 . We examined the expression levels of known genes involved in these two pathways and found that the HR pathway genes SMC6A, SMC6B 43 , RAD17, RAD51, RAD54 44 and REV7 45 , and the NHEJ pathway genes LIG4, KU70 and KU80 46,47 , were all up-regulated in the parg1-4 mutant compared to that in wild type after bleomycin treatment (Fig. 7A). These results suggested that the parg1 mutant cells might experience more DNA damage than wild type. To test this, we used the comet assay to observe DNA damages in wild type and mutant plants, where the proportion of the comet tail to the head represents the degree of DNA double-strand breaks in the nucleus 48 . No striking changes could be observed in the parg1-4 mutant under normal growth conditions. However, after treatment with bleomycin, the mutant showed a significantly higher level of DNA damage than wild type plants (Fig. 7B). When the genomic DNA was isolated and subjected to agarose electrophoresis, the DNA of parg1-4 migrated as a band with more smear than that of Col-0, indicating the presence of severe DNA damages in parg1-4 (Fig. 7C). Taken together, loss of function of PARG1 compromised the DNA repair process, although the DNA repair genes were still highly induced in the parg1 mutant.
Induction of PARG1 gene is ATM-and ATR-dependent and PARG1 represses the transcriptional up-regulation of ATM, ATR and SOG1. ATM and ATR are two critical kinases which transduce double and single strand break signals to DNA repair machinery, respectively 32 . They phosphorylate the transcription factor SOG1, which then induce the expression of DNA repair genes 33,34 . To understand if PARG1 participates in these two signaling pathways, we compared the expression level of PARG1 in Col-0, atm and atr mutants and found that, after bleomycin treatment, PARG1 expression levels in atm and atr mutants were obviously lower than that in wild type plants (Fig. 8A), indicating that ATM and ATR are necessary for the induction of PARG1. Surprisingly, when the transcript levels of ATM and ATR in Col-0 and parg1-4 were compared, ATM and ATR were both significantly induced in the parg1 mutant by bleomycin while not in wild type plants within the time periods we examined. The levels of ATM in the parg1 mutant at all time points were higher than those in wild type, except in the untreated plants (Fig. 8B). ATR was also induced constantly in the parg1 mutant, but the beginning level was lower than that of the wild type plants (Fig. 8C). When the expression level of the master transcription factor SOG1 was examined, it showed a similar up-regulation tendency in the parg1-4 mutant compared to Col-0

Discussion
There is a large family of PARP genes but only one PARG gene in human and mouse 25,26 . The single PARG gene fulfils different functions by generating multiple isoforms 24 . PARG gene knock out in mouse causes embryo lethality 25 , indicating an essential role of PARG for embryo development. In contrast, there are two PARG genes in Arabidopsis, and the parg1 and parg2 mutants as well as parg1parg2 RNAi transgenic plants all develop normally under standard growth conditions, suggesting that these genes are not critical for Arabidopsis normal development. However, under genotoxic stress, PARG1 becomes essential for plant survival. The two knock-out mutants, parg1-2 and parg1-4, are both hypersensitive to bleomycin. Loss of function of PARG1 leads to severe cell death in the mutant after bleomycin treatment, indicating a pivotal role of PARG1 in DNA DSB repair. However, only parg1-4 is highly sensitive to SSB inducing agent MMS and parg1-2 is not. This is probably due to that T-DNA in different sites interrupts the gene's function to different extents. In parg1-4, the T-DNA is integrated between Thr311 and Gly312 residues, which are just before the Tyr313 residue in the tyrosine clasp, and the tyrosine clasp is considered important for substrate binding 29 . Therefore, T-DNA in the parg1-4 mutant may destroy the substrate binding site of PARG1, thus inactivates the protein. In contrast, in the parg1-2 mutant, T-DNA is integrated between Ala408 and Ser409 residues, which is far from the PARG signature motif (G262 to E274) and also the tyrosine clasp. The parg1-2 mutant may preserve residual activity of PARG1, which is enough for SSB response, but not for DSB response. As for the PARG2 gene, both mutants are slightly sensitive to MMS or bleomycin. Inhibition of PARG2 enhanced the phenotype of the parg1-4 mutant to MMS, suggesting that PARG2 does play a role, although minor, in DNA repair. A recent report indicated that PARG2 is an inactive PARG enzyme 19 . However, the expression data that PARG2 is induced by genotoxin (Supplementary Fig. S3B) and knock-out of PARG1 leads to a compensatory up-regulation of PARG2 expression (Supplementary Fig. S3A), together with the phenotypes of parg2 single mutant and g1g2 interference plants, all strongly support a role of PARG2 in DNA repair. In mouse, the parp1-/-/parp2-/-double mutant, like the mouse parg mutant, is also embryo lethal 49 . However, the p1p2p3 triple mutant and g1g2 RNAi mutant in Arabidopsis both develop normally, suggesting that poly(ADP-ribosyl)ation, different from its role in animals, does not play an essential role in Arabidopsis development under standard conditions. Under genotoxic stresses, the phenotypes of the p1p2 and p1p2p3 mutants are also weak. Although PARPs can act as detectors for DNA double or single strand breaks, there are other known DNA damage sensors, such as DSB sensor MRN (MRE11/ RAD50/NBS1) complex and SSB sensors RPA protein and 9-1-1(RAD9/RAD1/HUS1) complex in plants 32,44 . Therefore, loss of functions of three PARPs does not cause disastrous consequence to plants. Furthermore, it is also considered that plants can bear more severe DNA damage than animals due to the immobile nature of cells and no risk for carcinogenesis 32 . By comparing the phenotypes of single, double and triple mutants of Arabidopsis PARP family, we conclude that PARP1 and PARP2 function in the DNA repair process at seedling stage, and PARP3 does not.
Transcriptional studies of the DNA repair genes in the parg1-4 mutant indicated that the repair pathway is activated, but not succeeded in repairing damage, resulting in the elevation of DNA damage level in the mutant. Over-accumulated DNA damage activates the cell death in the parg1 mutant, which firstly occurs in the meristems because the meristematic cells are undergoing active DNA replication, giving rise to the high sensitivity of these cells to DNA damaging agents 50 . PARG1 is primarily induced in root and shoot meristems, consistent with the important role of DNA repair in these cells. Besides cell death, DSBs can also induce endoreduplication in the root tips 51 , which is marked by an enlarged epidermal cell volume. Endoreduplication can prevent the damaged cells from propagation by undergoing DNA replication without cytokinesis. We observed a more pronounced epidermal cell size enlargement effect in the parg1-4 mutant (Fig. 5D), suggesting that the mutant also undergoes an enhanced reprogramming from mitosis to endocycle besides cell death.
Our expression data showed that PARG1 acts downstream of both ATM and ATR signaling pathways. ATM and ATR are the major kinases responsible for the activation of DNA repair pathways 32,38,44,52 . They regulate DNA repair process at the post-translational level by phosphorylation of the target proteins such as Chk1, Chk2 in animals and SOG1 in plants 32,33 . The induction of PARG1 is attenuated in both mutants, and the level is even lower in atm mutant, indicating a more pronounced effect by the absence of ATM. Interestingly, we found that the lack of PARG1 results in the expression level up-regulation of the key factors ATM, ATR and SOG1, and also other DNA damage responsive genes in plants after bleomycin treatment (Figs 7 and 8). In another word, expression of PARG1 somehow represses the induction of DNA damage responsive genes. It is recently identified in Arabidopsis that a point mutation of PARG1, which inactivates the PAR hydrolase activity, causes the overexpression of PR genes 19 . Therefore, PARG1 probably also regulates the amplitude of stress responses in addition to DNA damage response in plants. In animals, PARPs and PARGs can participate in stress granule formation and AGO (argonaute) proteins can be PARylated by PARPs, therefore relieves microRNA-mediated silencing of target genes 53 . In Arabidopsis, it remains unknown if disruption of PARG1 causes the PARylation level increase of AGOs under stress, thus alleviates the repression of stress-responsive gene by microRNA and leads to transcriptional up-regulation of these genes. Besides this, it is also possible that PAR acts as a stress signal in plants and induces a wide range of stress responses, including DNA damage response. PARG removes PAR so as to reduce the stress signaling. Poly(ADP-ribosyl)ation has been shown to be involved in biotic and abiotic responses in Arabidopsis 13,15,16,19,29,30 , but the molecular mechanisms are still not very clear until now.
PAR is a death signal in animal cells 54 , which promotes apoptosis-inducing factor (AIF) to translocate from mitochondrion to nucleus to initiate programmed cell death independent of caspase 55 . It is unknown whether this type of cell death exists in plants. The AIF homologue in Arabidopsis, MDAR1, is a monodehydroascorbate reductase mainly localized in peroxisome and involved in removing toxic H 2 O 2 . The MDAR genes are involved in multiple stress responses but the function in regulating cell death has not been demonstrated 56,57 . Our data indicated that an unknown cell death mechanism may be activated in the parg1 plants by PAR, since the parg1 mutant showed cell death phenotype after treatment with bleomycin, and a higher level of PAR is also detected in the mutant. Plants may recognize PAR as a stress signal when it is under a certain threshold but activate cell death when it exceeds the tolerance limit. A better understanding of how cell death is activated in the mutant would contribute to elucidating the molecular mechanism of cell death in plants.

RNA isolation and quantitative RT-PCR (qRT-PCR).
Total RNA was isolated using TRIzol Reagent (TaKaRa, Japan) and then quantified by measuring OD 260 . Complementary DNA (cDNA) was synthesized using PrimeScript RT-PCR Kit with gDNA eraser (Takara, Japan), and then used for qRT-PCR. PCR were performed with SyBR Premix Ex Taq TM II (TaKaRa, Japan) in a Real One Plus Real-Time PCR System (Applied Biosystem, USA), following the manufacturer's instructions. Specific primers used can be found in Supplementary Table S1. DNA constructs. For parg1-4 mutant complementation, a genomic DNA fragment of PARG1 was amplified using PCR (Primers are listed in Supplementary Table S2) and cloned into the Kpn I/Pst I sites of vector pZP221 58 .
For protein expression in E.coli, PARP1 coding sequence was cloned into pET32a vector (Novagen, USA) in frame with TRX tag using Sac I/Not I sites. PARG1 coding sequence was cloned into pGEX-4T-1 vector (GE Healthcare, UK) in frame with GST tag using EcoR I/Sal I sites (Primers are listed in Supplementary Table S2).
The RNAi construct of PARG2 was generated by amplifying the nucleotide sequence from 701 to 1140 bp downstream of the start codon of PARG2 gene, then cloning the fragment into pGEM-RNAi vector in inverted orientations to produce a hairpin fragment spanned by an intron, and finally the hairpin fragment was cloned into p35S-Fast vector 58 .
The construct of pPARG1::GUS was generated by amplifying 1140 bp fragment upstream of the start codon of PARG1 gene and then cloning the fragment into pAKK687 vector using Xba I and BamH I sites.
Protein expression and purification. TRX-PARP1 was produced in E.Coli Origami (DE3) (Novagen, USA) and GST-PARG1 was produced in Rosetta (DE3) (Novagen, USA). After OD 600 of the culture reached 0.6, 0.3 mM isopropyl-β -D-thiogalactopyranoside (IPTG) was added into the culture and incubated at 16 °C for 20 h. Bacteria were harvested by centrifuge and then broken by ultra-sonication. The soluble proteins were used for purification. GST and GST-PARG1 were purified using the Glutathione-Agarose affinity resin (Sigma Aldrich, USA) according to the manufacturer's instructions. TRX and TRX-PARP1 were purified using HisPur Ni-NTA resin (Thermo Scientific, USA).
Enzymatic activity assays. Preparation of auto-PARylated PARP1 substrate: Purified TRX-PARP1 and TRX proteins were dialyzed completely against 10 mM Tris·HCl pH 7.5 buffer to remove salts. 15 μ g TRX-PARP1 or 3 μ g TRX proteins were aliquoted into each tube containing 20 mM Tris·HCl pH 7.5, 50 mM NaCl, 7.5 mM MgCl 2 , 0.2 mM DTT, and 10 μ M fragmented DNA. To initiate the reaction, NAD + (Sigma Aldrich, USA) was added into each tube to a final concentration of 0.2 mM. The reactions were continued at room temperature for desired time periods and then terminated by adding 300 μ l pre-chilled acetone. After centrifugation, the precipitated proteins were subjected to SDS-PAGE and visualized by Coomassie blue staining. PARG1 activity assay on PARylated proteins: To remove contaminants from poly(ADP-ribosyl)ation reaction, the PARylated TRX-PARP1 substrate was dialyzed against pH7.0 phosphate buffer containing 1 mM 2-mercaptoethanol. Approximately 1 μ g PARylated TRX-PARP1 was incubated with 0.2 μ g GST or 0.5 μ g GST-PARG1 for different time periods in 100 μ l dialysis buffer, respectively. The reactions were terminated by the addition of 300 μ l ice cold acetone. After centrifugation, the proteins in each reaction were subjected to SDS-PAGE analysis and visualized by Coomassie blue staining. Immuno-blot analysis was performed following standard protocol using anti-PAR (Abcam, UK) or anti-PARP1 antibody (Shanghai Immune Biotech, China) as the primary antibody, respectively. Enhanced chemiluminescence reagents (Thermo Scientific, USA) were used for signal detection.
PARG1 activity assay on free PAR: 0.5 μ g GST-PARG1 or GST were incubated with 5 μ l of 10 μ M PAR (Trevigen, USA) substrate in reaction buffer (25 mM Tris pH 7.5, 150 mM NaCl) at a total volume of 20 μ l for different time periods. 2 μ l mixture was spotted on nitrocellulose membrane (GE Healthcare, UK). The membrane was allowed to dry at room temperature, then blocked with 1 X TBST plus 5% nonfat milk and detected with anti-PAR antibody.
Histochemistry observation. Leaf dead cell detection and H 2 O 2 staining were carried out as described previously 59 . For root dead cell detection, Arabidopsis seedling roots were stained with 10 μ g ml −1 propidium iodide (Sigma Aldrich, USA). Excitation and emission wavelength were set as 535 nm and 617 nm, respectively for fluorescence microscopy.
Chlorophyll determination and comet assay. The chlorophyll concentration of seedlings was measured as described 60 . The comet assay was performed based on the method reported before 48 .
Scientific RepoRts | 5:15892 | DOi: 10.1038/srep15892 DNA fragmentation assay. One week old seedlings were treated with 10 μ g ml −1 bleomycin for 48 hrs. Genomic DNA was extracted using the ChargeSwitch gDNA Plant Kit (Invitrogen, USA) and resolved in agarose gel. Each lane was loaded with 100 ng DNA.

Immuno-blot analysis of PAR in plants.
Arabidopsis seedlings were collected and ground in liquid nitrogen, then boiled in 1XSDS-PAGE loading buffer for 5 min and centrifuged. The supernatant was subjected to SDS-PAGE, transferred onto nitrocellulose membrane, and then detected with anti-PAR antibody. Equal loading was monitored by staining the membrane with Coomassie blue solution.