RUG3 and ATM synergistically regulate the alternative splicing of mitochondrial nad2 and the DNA damage response in Arabidopsis thaliana

The root apical meristem (RAM) determines both RAM activity and the growth of roots. Plant roots are constantly exposed to adverse environmental stresses that can cause DNA damage or cell cycle arrest in the RAM; however, the mechanism linking root meristematic activity and RAM size to the DNA damage response (DDR) is unclear. Here, we demonstrate that a loss of function in RCC1/UVR8/GEF-Like 3 (RUG3) substantially augmented the DDR and produced a cell cycle arrest in the RAM in rug3 mutant, leading to root growth retardation. Furthermore, the mutation of RUG3 caused increased intracellular reactive oxygen species (ROS) levels, and ROS scavengers improved the observed cell cycle arrest and reduced RAM activity level in rug3 plants. Most importantly, we detected a physical interaction between RUG3 and ataxia telangiectasia mutated (ATM), a key regulator of the DDR, suggesting that they synergistically modulated the alternative splicing of nad2. Our findings reveal a novel synergistic effect of RUG3 and ATM on the regulation of mitochondrial function, redox homeostasis, and the DDR in the RAM, and outline a protective mechanism for DNA damage repair and the restoration of mitochondrial function that involves RUG3-mediated mitochondrial retrograde signaling and the activation of an ATM-mediated DDR pathway.


Results
A loss of function in RUG3 results in short roots and delayed development. To identify novel genetic loci that regulate root growth, we performed a phenotypic analysis of loss-of-function mutants of 24 genes encoding RCC1 family proteins by sowing the seeds on Murashige and Skoog (MS) medium for 8 days. The mutants with mutations in RUG3 (At5g60870) caused by T-DNA insertions ( Supplementary Fig. S1) and undetectable levels of RUG3 mRNA showed severe root growth defects (Fig. 1a-c). The mutants (SALK_092071 and N336814) we used in this study were previously designated as rug3-1 and rug3-2, respectively 30 .
A previous study showed that a loss of function in RUG3 caused retarded development in Arabidopsis including the root growth defects 30 ; however, a detailed description of the influence of RUG3 mutations on root growth and development is lacking. To further investigate the role of RUG3 in plant root development, we conducted a systematic phenotypic analysis of rug3 mutants. As shown in Fig. 1c-h, rug3-1 and rug3-2 mutant plants grew much more slowly than did wild-type (WT) roots after radical emergence, and the primary root length in both rug3 mutants was reduced by approximately 55% at 8 days after stratification (Fig. 1c,d). Notably, the size of the RAM in the mutant primary roots was markedly reduced, and the average cortical cell number in the RAM of the rug3 plants was dramatically decreased compared with that in wild type (Fig. 1e,f). Furthermore, the average length of mature root cells in the rug3 mutants was dramatically decreased compared with that in wild type (Fig. 1g,h). These results suggest that RUG3 modulates both RAM activity and root cell elongation in Arabidopsis.
In addition to root growth, both rug3 mutants showed delayed development. For example, the mutants exhibited late germination and cotyledon greening compared with WT plants (Supplementary Fig. S2a-c). At the seedling stage, the rug3 plants were much smaller than wild type ( Supplementary Fig. S2d), and the flowering time of the mutants was slightly delayed (Supplementary Fig. S2e; Supplementary Fig. S3a,b). As a result, the life span of the mutants was dramatically extended compared with that of wild type. Also, the stature of the rug3 mutants was comparable to that of wild type at the mature stage. Together, these data indicate that RUG3 also plays a pivotal role in vegetative growth and the phase transition from vegetative to reproductive growth. RT-PCR analysis of the expression of RUG3 in wild type, rug3-1, and rug3-2. ACT2 was used as an internal control. The full-length gel can be find in supplementary information file the full-length gel for (b). (c) Phenotypes of WT, rug3-1, and rug3-2 seedlings at 8 days after germination (DAG). Three-day-old seedlings grown on MS medium were grown vertically on MS medium for another 5 days before being photographed. Bar = 1 cm. (d) Quantitative analysis of primary root length at different time points. Three biological repeats were done with similar results. The data shown are means ± standard deviation (SD) (n = 30). (e) Cortical cell number in the meristematic zone in WT, rug3-1 and rug3-2 plants at 6 DAG. The data shown are means ± SD (n = 30). Different letters represent significant differences. (f) Comparison of root meristem size in WT, rug3-1 and rug3-2 plants at 6 DAG. The vertical lines indicate the root meristem length. Bar = 100 μ m. (g) Mature root cell length in WT, rug3-1 and rug3-2 plants at 6 DAG. The images were taken after Propidium Iodide (PI) staining. Bar = 60 μ m. (h) Quantitative analysis of the mature root cell lengths in WT, rug3-1 and rug3-2 young seedlings at 6 DAG. The data shown are means + SD (n = 30). Different letters represent significant differences (Student-Newman-Kuels test, P < 0.05).
Since these mutant alleles of RUG3 produced the same developmental defects, we assessed whether the mutant phenotypes were caused by a loss of function in RUG3. As expected, overexpression of RUG3 under the control of the CaMV35S promoter in a rug3-1 background completely restored all of the developmental phenotypes of the rug3-1 mutant ( Supplementary Fig. S4a-e), confirming the aforementioned role of RUG3 in plant development. Thereafter, we used rug3-1 to examine the function of RUG3 in root growth and plant development.
The rug3 mutant show increased cell cycle arrest at the G2/M checkpoint. To assess whether the short roots observed in rug3-1 plants were due to a defect in cell cycle regulation, we analyzed the expression of the marker gene of cell cycle in rug3-1 mutant plants. Our expression analysis show that the expression of KRP2, which is a main negative regulator of G1/S checkpoint 31 , was not changed ( Supplementary Fig. S4), and the expression of CDKB1;1, which is specifically activated during early S phase and M phase 32 was also not dramatically altered in rug3-1 mutant plants ( Fig. 2a; Supplementary Fig. S5). By contrast, the expression of CycB1;1, WEE1 and CDKB2;1, which activate the G2/M checkpoint 19,[33][34][35] are upregulated ( Fig. 2a; Supplementary Fig. S5). To confirm the result, we generated rug3-1 lines expressing WEE1pro::GUS and CycB1;1pro::GUS assays confirmed the upregulation of WEE1 and CycB1;1 transcription in the rug3-1 mutant background. GUS assays confirmed the upregulation of WEE1 transcription in the rug3-1 mutant background (Fig. 2b). The GUS assays also revealed CycB1;1pro::GUS expression in the meristematic tissues of germinating WT seedlings; in sharp contrast, CycB1;1pro::GUS expression was substantially increased in rug3-1 roots and shoots during early development (Fig. 2c,d). Taken together, these results indicate that RUG3 may function as a negative regulator of G2/M checkpoint activation.
The mutation of RUG3 exacerbates DNA damage. The G2/M checkpoint is also known as the G2/M DNA damage checkpoint, ensures that cells repair damage to their DNA before entering mitosis 19,36 . Based on the role of RUG3 in controlling the G2/M checkpoint, we speculated that RUG3 is involved in regulating the DDR in Arabidopsis. If this were the case, rug3-1 mutants should exhibit excessive DNA damage. To test this, we performed a comet assay to evaluate the DNA damage in 9-day-old seedlings grown under normal conditions. As expected, the percentage of DNA in the tail of rug3-1 cells was significantly higher than that in wild type (Fig. 3a,b), pointing to a role for RUG3 in the DDR at the G2/M checkpoint. Next, we examined the response of rug3 plants to the DSB-inducing agent methyl methanesulfonate (MMS; 0.01%) and found that rug3-1 was hypersensitive to MMS (Fig. 3c). Treated rug3-1 seedlings displayed greater root and shoot growth inhibition than did WT seedlings (Fig. 3c,d). These results suggest an important role for RUG3 in the DSB repair response. To further assess whether RUG3 mediates DNA replication, we analyzed the sensitivity of rug3-1 plants to a chemical inducer of DNA damage, the ribonucleotide reductase inhibitor hydroxyurea (HU). The result shows that the sensitivity of the rug3-1 plants to HU was comparable to that of the wild type ( Supplementary Fig. S6), indicating that the RUG3 does not mediate the DNA replication stress in Arabidopsis.
Since the rug3-1 mutant show hypersensitive to DSB-inducing agent MMS, we assessed whether the mutant phenotypes were caused by a loss of function in RUG3. As expected, the transgenic line Re-4 can completely restored the greening rate phenotypes of the rug3-1 mutant and partial restoration the short root phenotypes ( Supplementary Fig. S7), confirming the aforementioned role of RUG3 in DNA damage response.
RUG3 is ubiquitously expressed and may be repressed by MMS. Although it has been reported that RUG3 is a mitochondrial protein based on the bioinformatic prediction of an N-terminal transit peptide targeting the protein to mitochondria and the demonstrated subcellular localization of N-terminus RUG3 -GFP in Arabidopsis protoplasts 30 , information about the expression pattern of RUG3 and the subcellular localization of RUG3 is lacking. To address this, we first analyzed the expression pattern of RUG3. Semi-quantitative PCR showed that RUG3 was expressed in multiple tissues, including the roots and rosette leaves of young seedlings and in the flowers and siliques of mature plants, with higher expression levels in the roots and rosette leaves of young seedlings (Fig. 4a). We next generated transgenic lines expressing RUG3pro::GUS for histochemical staining; our results confirmed the ubiquitous expression of RUG3. RUG3 was preferentially expressed in the vascular tissues of the various organs and tissues tested (Fig. 4b-e), the preferential expression in the vasculature fits to the fact that phloem cells contain most mitochondria 37 .
To check whether RUG3 is responsive to DNA damage and DNA replication stress, we analyzed the transcription level of RUG3 in response to MMS treatment. As shown in Fig. 4f, the mRNA level of RUG3 was markedly reduced by MMS. GUS staining confirmed that RUG3 transcription was repressed by MMS (Fig. 4g).

RUG3 is localized to mitochondria and interact with ATM.
To further characterize the subcellular localization of full-length RUG3, we expressed RUG3-GFP under the control of the 35 S promoter in Nicotiana benthamiana leaf cells. As shown in Fig 5a, RUG3-GFP was detected in mitochondria. The observed mitochondrial localization of RUG3 is consistent with previous data obtained using the N-terminus of RUG3 30 , and it confirms the role of RUG3 in modulating mitochondrial function. RCC1 is a guanine nucleotide exchange factor (GEF) for the small GTP-binding protein Ran 38,39 . Although we find RUG3 is localized to mitochondria, we still wondered whether RUG3 has the similar roles. As we expected, the results showed that, like UVR8 (UV RESISTANCE LOCUS 8), RUG3 did not have detectable GEF activity ( Supplementary Fig. S8).
ATM is a key upstream regulator of the DDR, and it was recently demonstrated to localize to and function in mitochondria in human cells 40,41 . ATM is a large protein composed of 3845 amino acids; thus, it was difficult to express the full-length protein. Since ATM contains an N-terminal PWWP domain, which is involved in protein-protein interactions 42 , we used the N-terminal region of ATM (300 amino acids; ATMn) in our experiments. To further characterize the subcellular localization of Arabidopsis ATM, we expressed the N-terminal sequence of ATM fused with GFP under the control of the 35 S promoter in Nicotiana benthamiana leaf cells. Localization of ATMn-GFP in mitochondria was confirmed by Mitotracker staining (Supplementary Fig. S9).
Because RUG3 is colocalized with ATM in mitochondria and has a similar role in the DDR, we speculated that RUG3 and ATM regulate mitochondrial function and cellular responses to DNA damage via a direct or indirect interaction. To test this possibility, we performed protein-protein interaction assays. Intriguingly, a bimolecular fluorescence complementation (BiFC) assay showed that RUG3 interacted directly with ATMn in N. benthamiana leaf cells (Fig. 5b). This interaction was confirmed by a pull-down assay (Fig. 5c). Thus, RUG3 and ATM may work together to modulate mitochondrial function and the DDR. ATM and RUG3 synergistically regulate the DNA damage stress response. ATM is a key upstream regulator of the DDR, and then we analyzed the genetic relationship between RUG3 and ATM in response to DNA DSBs in atm-2rug3-1 double mutant plants produced by a genetic cross. The sensitivity of rug3-1, atm-2 43 , and atm-2rug3-1 double mutant plants to MMS stress was evaluated. When germinated on medium supplemented with 0.01% MMS, rug3-1 and atm-2 showed greater sensitivity than WT plants to MMS treatment (Fig. 6a,b; Supplementary Fig. S10). The percentage of germinating rug3-1 or atm-2 seedlings that turned green was markedly reduced by about 50% (Fig. 6b). By sharp contrast, MMS sensitivity was exacerbated in germinating atm-2rug3-1 double mutant plants, and the greening rate dropped sharply to only 12% (Fig. 6b). It is possible that RUG3 and ATM play a synergistic role in regulating the DDR.
ATM enhances the regulation of nad2 splicing by RUG3. It has been reported that RUG3 regulates nad2 pre-mRNA splicing 30 . Given that ATM and RUG3 synergistically regulate the DDR in plants, we speculated that ATM co-regulates the alternative splicing of nad2 pre-mRNA. If this is the case, atm-2 mutant plants should have abnormal nad2 splicing efficiency and atm-2rug3-1 mutant plants should exhibit even greater defects in nad2 splicing. To test this, we analyzed the level of nad2 mRNA containing intron 3 in rug3-1, atm-2, and atm-2rug3-1 double mutant plants following MMS (0.01%) treatment. As shown in Fig. 6c, the amount of nad2 mRNA containing intron 3 was significantly increased in rug3-1 mutant, consistent with previous data 30 . Unexpectedly, nad2 splicing in the atm-2 mutant was comparable to that in wild type. However, to our surprise, the retention of intron 3 in nad2 mRNA was elevated in the atm-2rug3-1 double mutant plants. To rule out the possibility that increased retention of intron 3 in nad2 mRNA is derived from an overall increase in mitochondrial transcripts  of nad2, we performed RT-PCR to detect both the unspliced and spliced nad2 transcripts under MMS treatment and calculated the ratio of unspliced transcripts to spliced transcripts. As expected, the unspliced and spliced nad2 were increased and decreased in rug3-1 and atm-2rug3-1 double mutant plants, respectively (Fig. 6d). As a result, the ratio of unspliced transcript to spliced transcript in atm-2rug3-1 double mutant is significantly higher than rug3-1 (Fig. 6e). These data indicate that ATM and RUG3 synergistically regulate alternative splicing of nad2 pre-mRNA.
A loss of function in both RUG3 and ATM causes ROS overproduction. The gene nad2 encodes a subunit of mitochondrial complex I, which is the main source of mROS. Since RUG3 and ATM synergistically regulate nad2 splicing and the DDR, we hypothesized that the increased DDR and cell cycle arrest in the RAM may be due to elevated ROS levels caused by abnormal nad2 splicing and complex I functionality. We therefore analyzed the role of RUG3 in modulating ROS levels and the ROS responses of plants. 3,3′ -Diaminobenzidine (DAB) staining showed that 7-day-old rug3-1 mutant plants had substantially increased H 2 O 2 levels in the RAM and cotyledons under normal conditions compared to wild type (Fig. 7a,b). The high level of ROS accumulation in rug3-1 was confirmed by dichloro-dihydro-fluorescein diacetate (DCFH-DA) staining ( Supplementary Fig. S11), suggesting a role for RUG3 in maintaining ROS homeostasis.
To assess whether the elevated ROS level in the rug3-1 mutant causes the DDR and cell cycle arrest in the RAM, we tested the effect of a scavenging agent on the ROS content in rug3-1 mutant plants. DAB staining clearly showed that the addition of GSH to the growth medium (MS medium) almost completely reversed the high level of ROS in the RAM and largely restored the RAM phenotypes of rug3-1 mutant plants (Fig. 7c,d). We also used the ROS production inhibitor diphenyleneiodonium (DPI) to analyze that whether reduced ROS production in mitochondria would release the cell cycle arrest defect seen in rug3. Following treatment with DPI for 10 h, the CycB1;1pro::GUS signal in rug3 was substantially reduced to the level in wild type (Fig. 7e). These data confirm that RUG3 modulates cell cycle progression in the RAM by maintaining ROS homeostasis.
Next, we tested whether RUG3 and ATM co-regulate mROS accumulation in response to MMS treatment. As shown in Fig. 7f and g, in the presence of MMS, atm-2 plants accumulated more H 2 O 2 than did wild type. The level of H 2 O 2 in the rug3-1 mutants was higher than that in wild type and increased significantly in response to MMS treatment. The level of H 2 O 2 in the atm-2rug3-1 double mutant plants was similar to that in the rug3-1 mutant with or without MMS treatment. However, further phenotypic analysis showed that the two single mutants and double mutant exhibited different levels of sensitivity to H 2 O 2 treatment compared with wild type (Supplementary Fig. S12); atm-2 showed the least sensitivity to H 2 O 2 , rug3-1 was in the middle, and the atm-2rug3-1 double mutant showed the greatest sensitivity to H 2 O 2 . Together, these results suggest that ATM does not affect ROS accumulation in plants under normal conditions, but that it modulates ROS accumulation and sensitivity by working together with RUG3 in the DDR.

Discussion
The growth and development of plants, which are sessile organisms, occur mainly postembryonically and are affected by both developmental and environmental factors. One of the most important questions in plant biology is how meristematic activity and meristem tissue homeostasis are controlled in apical meristems, which determine shoot and root growth rates and the overall architecture of plants. In the RAM, how the meristematic activity of stem cells and their daughter cells is regulated is not precisely known. In particular, the mechanism by which RAM size and activity are modulated in response to the DNA damage caused by abiotic stress remains elusive. RUG3 was previously shown to regulate nad2 pre-mRNA splicing and complex I biogenesis 30 . In the present study, we found that RUG3 is an upstream regulator of ROS, which trigger retrograde signaling from mitochondria to the nucleus in response to oxidative stress and regulate both the DDR and cell cycle progression in the RAM. We also uncovered a novel mechanism by which ATM interacts with RUG3 in mitochondria to synergistically regulate the alternative splicing of nad2 pre-mRNA, mitochondrial function, oxidative metabolism, the DDR, and RAM activity.
In human cells, RCC1 acts as a guanine nucleotide exchange factor (GEF) for Ran GTPase; it also has chromatin and DNA binding activity in the nucleus, enabling it to regulate gene expression 38,39,44 . In plants, there are 24 genes encoding RCC1 family proteins; among them, UVR8 and TCF1 (Tolerant to Chilling and Freezing 1) have been well characterized and both proteins mediate plant response to environmental conditions [45][46][47][48][49] . In cells irradiated with ultraviolet B (UVB) light, UVR8 moves from the cytosol to the nucleus, where it activates a signaling cascade that enables the cell to respond to the exposure 46 . When plants are exposed low temperature, TCF1 modulates lignin contents of the stressed cells and plant freezing tolerance through a chromatic-based regulatory network 49 . RUG3 encodes an RCC1 family protein that contains seven RCC1 tandem repeats and has strong similarity to UVR8. Unexpectedly, RUG3 contains mitochondrial targeting sequences (MTS); both a partial protein containing the MTS 30 and full-length RUG3 driven by the native promoter were shown to localize to mitochondria (Fig. 5a), suggesting a novel role for RUG3 in mitochondrial function. Indeed, it has previously been shown that RUG3 is specifically required for complex I biogenesis; it regulates the splicing of intron 3 from nad2 pre-mRNA molecules 30 . Our results also show that, like UVR8, RUG3 does not have detectable GEF activity (Supplementary Fig. S8). Since no DNA or RNA binding activity has been detected for RUG3, it is unclear how RUG3 mediates nad2 pre-mRNA splicing. It may be involved in nad2 pre-mRNA splicing through an interaction with ABO5, a PPR protein that can bind RNA and which is required for the splicing of intron 3 from nad2 pre-mRNA 28 , or through an interaction with other proteins (e.g., mitochondrial transcription termination factor 15 [mTERF15]) that specifically modulate intron 3 splicing from nad2 pre-mRNA 50 .
A previous study showed that a loss of function in RUG3 resulted in global changes in transcription, including the transcription of genes that function in mitochondria, protein metabolism, and cell organization 30   evidence to support the hypothesis that RUG3 is an upstream regulator of mitochondrial retrograde signaling and that it controls cell cycle progression in the RAM by modulating ROS production and the DDR. First, we showed that RUG3 is a positive regulator of meristematic activity in the RAM, and that a loss of function in RUG3 resulted in short roots (Fig. 1c-h). Second, we demonstrated that RUG3 is required for cell cycle progression, and that the mutation of RUG3 caused a cell cycle arrest (Fig. 2). Third, we demonstrated that the cell cycle arrest mediated by RUG3 was due to activation of the DDR. The mutation of RUG3 resulted in the increased expression of DDR markers (e.g., BRCA1 and RAD51) and downstream effectors (e.g., WEE1) that control the G2/M cell cycle arrest checkpoint (Supplementary Fig. S5; Supplementary Fig. S13). Finally, we showed that RUG3 plays an important role in ROS metabolism. Nonspecific and targeted scavenging of ROS restored the ROS levels and cell cycle progression in a loss-of-function RUG3 mutant. This is not surprising because mitochondrial complex I is the main site of ROS production 8,28,29 . The overproduction of ROS can cause DNA oxidation and damage 40,51 ; however, mROS are a key mitochondrial signal that triggers mitochondrial retrograde signaling, which reprograms the expression of nuclear genes and maintains mitochondrial function (Chandel 2014). Notably, previous studies showed that RUG3 regulates the expression of several genes encoding proteins that are essential for mitochondrial function (e.g., TOM6, TIM13, and TIM22) 30 . These data suggest a critical role for RUG3 in the regulation of retrograde signaling from mitochondria to the nucleus and in the maintenance of normal mitochondrial functioning. In addition to transcriptional changes, ROS may cause the oxidation of proteins associated with many other biological processes. For example, oxidative stress induces the oxidation of ATM, which is able to activate ATM in the absence of DNA damage 52 .
In mammals, ATM not only activates the DDR and cell cycle arrest, it also modulates mROS levels and mitochondrial functioning 40,41 . It has been shown that in mitochondria ATM phosphorylates BH3-interacting domain death agonist (BID), which induces oxidative stress and a cell cycle arrest, as well as the quiescence/renewal of hematopoietic stem cells 41 . However, there are no reports on the mitochondrial localization or function of ATM in plants. In this study, we preliminary found that the ATM is able to interact physically with RUG3 in mitochondria (Fig. 5b,c). Intriguingly, ATM and RUG3 had a synergistic effect on nad2 pre-mRNA alternative splicing ( Fig. 6c-e) and subsequent ROS homeostasis (Fig. 7f,g) and the DDR (Fig. 6a,b). Our data suggest that RUG3 and ATM function as a rheostat to maintain ROS homeostasis by regulating nad2 pre-mRNA alternative splicing, which is required for proper mitochondrial organization and complex I activity. In this model, RUG3 promotes normal nad2 pre-mRNA alternative splicing to ensure normal mitochondrial architecture and basal ROS levels under normal conditions; in response to the DDR, the repression of RUG3 results in decreased splicing efficiency, mitochondrial complex I dysfunction, and increased ROS levels. ROS, as mitochondrial retrograde signals, enter the nucleus to activate the ATM-dependent DDR. ATM is also translocated to mitochondria where it attenuates the abnormal alternative splicing of nad2 pre-mRNA by complexing with RUG3, and in doing so it helps maintain mitochondrial function and ROS homeostasis. To the best of our knowledge, this is the first report on the role of ATM in mitochondrial nad2 pre-mRNA splicing. However, the mechanism underlying ATM-mediated nad2 pre-mRNA splicing remains unclear. Recently, it was demonstrated in human cells that ATM targets the core spliceosome in response to DNA damage to regulate genome-wide alternative splicing with a preferential role in intron retention during the DDR 53 . It will be of great interest to determine whether ATM specifically regulates nad2 pre-mRNA splicing or whether it regulates the alternative splicing of multiple mitochondrial pre-mRNAs in response to DNA damage. Defining the exact mechanism underlying the ATM-and RUG3-mediated maintenance of mitochondrial function and ROS homeostasis in response to DNA damage will further enhance our understanding of the DDR and the regulation of tissue homeostasis in the RAM.

Methods
Plant materials and growth conditions. Arabidopsis thaliana ecotype Columbia-0 was used in this study.
Seeds were surface-sterilized with 50% (v/v) commercial bleach for 5 min, followed by five rinses with sterilized water, and then grown on MS medium (M519; PhytoTechnology Laboratories, Shawnee Mission, KS) containing 1% (w/v) sucrose and 0.8% agar. After 2 days at 4 °C, the seeds were germinated at 23 °C under a 16/8 h light/dark photoperiod.
The marker lines and mutants used in this work are rug3-1, rug3-2, atm-2, CycB1;1pro::GUS and Gene expression analysis. Total RNA was extracted from plant samples using Trizol reagent (Tiangen Biotech [Beijing] Co. Ltd., Beijing, China). First-strand cDNA was synthesized from the total RNA using a FastQuant RT Kit (TransGen Biotech Co., Ltd., Beijing, China). Semi-quantitative RT-PCR and qRT-PCR were done as described previously 5 . ACT2 was used as an internal control; three independent replicates were performed per experiment. The gene-specific primers used for WEE1, CycB1;1, CDKB2;1, and NAD2-intro3 are listed in Supplementary Table S1.
Analysis of nad2 transcript splicing. Measurement of the expression of intron 3 of nad2 and the splicing of nad2 by Semi-quantitative RT-PCR in plants treated with MMS. Fifteen-day-old seedlings grown on MS medium containing 0.01% MMS were used for analysis. The relative grey intensities normalized with UBC were calculated by the Image J software. The gene-specific primers used are listed in Supplementary Table S1. GUS staining assay. The promoter of RUG3 was cloned into pCAMBIA1391 at the EcoRI-BamHI sites to generate the construct containing RUG3pro::GUS. RUG3pro::GUS was then transformed into Arabidopsis by the floral dip method. Histochemical staining was performed as described previously 49 . construction, the coding sequence of RUG3 was amplified by PCR and the product was digested for insertion into the binary vector pCAMBIA1300 at the XbaI-BamHI sites. For RUG3pro::RUG3-GFP construction, the promoter of RUG3 (1416 bp) and the coding region of RUG3 were amplified by PCR, and the products were digested with PstI-XbaI and XbaI-BamHI, respectively, and then were inserted into the binary vector (pCAMBIA1300) sequentially. For 35 S::ATMn-GFP construction, the coding sequence of the ATM N-terminal sequence (300 amino acids; ATMn) amplified by PCR was cloned to pGW-B6 through the gateway system. The gene-specific primers used for the construction are listed in Supplementary Table S1. GEF activity assays. GEF activity assay was performed as described in 54 . Briefly, RCC1, RUG3, and human Ran were expressed in Escherichia coli as fusions with GST. The Ran clone was provided by Dr. Murray Stewart (Medical Research Council Laboratory for Molecular Biology, Cambridge, U.K.). Assays of guanine nucleotide exchange activity were performed by using [ 3 H]GDP to load 30 pmol GST-Ran and subsequent incubation with 0.5 nM recombinant RCC1 or RUG3 for 3 min. The exchange activity was calculated. The GEF assays were repeated four times.

Construction of
Yeast two-hybrid assays. The RUG3 coding region and RANs coding region were amplified by PCR and cloned in pGADT7 and pGBKT7 respectively. Plasmid DNA of bait and prey constructs were transformed into the S. cerevisiae strain Y190. Individual transformants were streaked on plates containing a synthetic, minimal (SD) medium lacking tryptophan and leucine and grown for 24 h. Yeast cells were transferred onto a filter paper, and β -galactosidase (β -gal) filter assays were performed as described by Zhu et al. 55 .

Measurement of ROS in plants.
For the DCFH-DA staining assay, root tips from 7-day-old seedlings were incubated in darkness for 30 min in a solution containing phosphate-buffered saline (PBS; pH 6.0) and 50 μ M DCFH-DA. They were then washed three times with PBS (pH 6.0) to remove the excess DCFH-DA. Fluorescence was detected with a Leica SP8 confocal microscope.
An Amplex Red Hydrogen Peroxide/Peroxidase Assay Kit (Invitrogen) was used to measure H 2 O 2 production as described by Shin et al. 56 .
For DAB staining to detect H 2 O 2 , seven-day-old seedlings were incubated in 0.1 mg/mL DAB (Sigma-Aldrich Corp., St. Louis, MO) dissolved in ddH 2 O (pH 5.0) for 8 h, and then in 75% ethanol at 65 °C for 15 min; after three washes, the roots were photographed.
Comet assay. A comet assay was performed using a Comet Assay Kit (4250-050-K; Trevigen, Gaithersburg, MD). The comets were visualized by staining with SYBR Green I, and then photographed using a Leica SP8 confocal microscope. Each experimental point is represented by the mean value from three independent experiments. For each material at least 50 nucleuses were photographed, then the data were analyzed by using Casp_1.2.3b1 software.
The coding sequence of RUG3 was cloned into pEarlygate201-YN to generate RUG3-YFP N while the ATM N-terminus was cloned into pEarlygate202-YC to generate ATM-N-YFP C . Agrobacterium GV3101 cells containing each gene were cultured overnight, and then adjusted the OD 600 at 0.6-0.8 using 10 mM MgCl 2 . An equal volume of each culture was mixed together for injection. At 2 days after injection, YFP fluorescence was detected with a Leica SP8 confocal microscope.
Pull-down assays. RUG3 was cloned into pGEX-4T-1 at the Sa1I-BamHI sites to generate GST-RUG3 fusion protein, while the ATM N-terminus (ATM-N, 1-900 bp) was cloned into pMAL-c2X at the EcoRI-BamHI sites to generate MBP-ATM-N. GST-or MBP-tagged proteins were expressed in Escherichia coli BL21 cells. For the pull-down assays, equal amounts of purified MBP-ATM-N were added to tubes containing GSH-Sepharose resin (L008; TransGen Biotech Co., Ltd.) plus GST or GST-RUG3 protein, respectively. Anti-MBP antibodies (E8030; New England Biolabs, Ipswich, MA) were used to detect the interaction between RUG3 and ATM-N.