ATM Promotes RAD51-Mediated Meiotic DSB Repair by Inter-Sister-Chromatid Recombination in Arabidopsis

Meiotic recombination ensures accurate homologous chromosome segregation during meiosis and generates novel allelic combinations among gametes. During meiosis, DNA double strand breaks (DSBs) are generated to facilitate recombination. To maintain genome integrity, meiotic DSBs must be repaired using appropriate DNA templates. Although the DNA damage response protein kinase Ataxia-telangiectasia mutated (ATM) has been shown to be involved in meiotic recombination in Arabidopsis, its mechanistic role is still unclear. In this study, we performed cytological analysis in Arabidopsis atm mutant, we show that there are fewer γH2AX foci, but more RAD51 and DMC1 foci on atm meiotic chromosomes. Furthermore, we observed an increase in meiotic Type I crossovers (COs) in atm. Our genetic analysis shows that the meiotic phenotype of atm rad51 double mutants is similar to the rad51 single mutant. Whereas, the atm dmc1 double mutant has a more severe chromosome fragmentation phenotype compared to both single mutants, suggesting that ATM functions in concert with RAD51, but in parallel to DMC1. Lastly, we show that atm asy1 double mutants also have more severe meiotic recombination defects. These data lead us to propose a model wherein ATM promotes RAD51-mediated meiotic DSB repair by inter-sister-chromatid (IS) recombination in Arabidopsis.


INTRODUCTION
Meiosis is a fundamental biological process during sexual reproduction in eukaryotes that generates haploid gametes (sperms and eggs) in preparation for fertilization. To successfully complete meiosis, a series of homologous chromosome interactions are required, including paring, synapsis and meiotic recombination (Page and Hawley, 2003). Meiotic recombination, specifically crossovers (COs), ensures accurate homologous chromosome segregation during the first meiotic division. Meiotic recombination is initiated by the formation of DNA double strand breaks (DSBs) which are catalyzed by the conserved topoisomerase-like protein SPO11 (Grelon et al., 2001;Stacey et al., 2006). These DSBs are processed by the MRN (MRE11-RAD50-NBS1) complex to create 3 single-stranded DNA (ssDNA) overhangs (Bleuyard et al., 2004;Waterworth et al., 2007;Samanic et al., 2013). The ssDNA is bound by two homologs of bacterial RecA protein, RAD51 and DMC1, forming nucleoprotein filaments that facilitate searching for homologous DNA templates for repair (Vignard et al., 2007;Goldfarb and Lichten, 2010). During meiosis, DSB repair preferentially utilizes homologous chromosomes as repair templates to produce COs or non-crossovers (NCOs) Shrivastav et al., 2008). The number and distribution of meiotic COs per meiocyte is regulated to ensure accurate homologous chromosome segregation (Jones and Franklin, 2006). However, the number of DSBs usually substantially exceeds the number of COs in each meiosis Serrentino and Borde, 2012). There are ∼150-250 DSBs in each Arabidopsis meiosis, but only ∼9-10 COs (Copenhaver et al., 1998;Serrentino and Borde, 2012;Choi et al., 2018). Arabidopsis has at least two types of COs: Type I and Type II (Copenhaver et al., 2002). Type I COs, which comprise ∼85% of total COs, are interference sensitive and depend on ZMM proteins (MLH1,HEI10,et al.); whereas Type II COs (∼15%) are interference insensitive and rely on MUS81 (Copenhaver et al., 2002;Mercier et al., 2015). CO interference is a poorly understood phenomenon that prevents closely spaced double COs (Berchowitz and Copenhaver, 2010). DSBs that are not repaired as COs are repaired by inter-homolog recombination as NCOs or inter-sister-chromatid recombination (Lu et al., 2012;Wijnker et al., 2013).
The DNA damage response kinase, Ataxia-telangiectasia mutated (ATM) is a central regulator of DNA damage response pathways. DSBs signals activate ATM which in turn phosphorylates numerous effector proteins for DNA damage responses in somatic cells (Marechal and Zou, 2013). ATM also has multiple roles in meiosis (Shiloh and Ziv, 2013;Udayakumar et al., 2015). In yeast and mouse, ATM negatively regulates meiotic DSB formation and distribution (Lange et al., 2011;Mohibullah and Keeney, 2017). ATM can also suppress the formation of extra meiotic COs (Barchi et al., 2008;Anderson et al., 2015). Interestingly, in mouse, ATM deficiency results in an increase of Type I COs (Barchi et al., 2008). In contrast, the deletion of the yeast ATM homolog Tel1 results in an increase of Type II COs (Anderson et al., 2015). ATM is also required to implement the meiotic checkpoint response in yeast (Penedos et al., 2015) and for phosphorylation of the DNA damage related histone H2AX during meiosis in mouse and yeast (Stucki and Jackson, 2006;Widger et al., 2018). However, whether or not Arabidopsis ATM also has similar meiotic functions is unclear.

Isolation and Identification of a Novel atm Mutant
To identify genes that regulate meiotic recombination, we screened a Ds insertion library generated in Landsberg erecta (Ler) Arabidopsis (Sundaresan et al., 1995), seeking sterile plants with meiotic defects. We found a sterile line, designated 184. The mutant has normal vegetative growth, but shorter siliques compared with that of wild type ( Figure 1A). In contrast to wild type flowers, the mutant does not shed pollen grains onto its stigma ( Figure 1B). Alexander staining revealed that about 90% pollen (viable/total: 41/445, n = 34) is non-viable in the mutant ( Figure 1C). Light microscope imaging of Toluidine Blue-stained tetrad-stage meiocytes showed that 82% of meiocytes in the mutant produce polyads after meiosis, compared to the four microspores observed in wild type (Figure 1D), indicating a defect in male meiosis. We crossed heterozygous line 184 (female) with wild type to produce an F1 heterozygote which we then allowed to set self-fertilized F2 seed. The F2 progeny plants had a ∼3:1 segregation ratio for fertile (104) and sterile (36) individuals (P = 0.89, chi-square test), indicating that the mutation is a recessive allele.
To determine whether the male meiotic defects in the mutant are associated with aberrant chromosome morphologies, we stained chromosome spreads with 4 , 6-diamidino-2phenylindole (DAPI). Chromosomes from wild type and mutant meiocytes exhibit no obvious morphological differences up to diakinesis ( Figure 1E). Similar to former reported atm-1 mutant, we did not observe any univalents in diakinesis cells (n = 51) of atm-5 (Garcia et al., 2003). At metaphase I, 29% of mutant meiocytes (n = 38) have two early separated homologous chromosomes, whereas wild type metaphase I meiocytes consistently have five bivalents ( Figure 1E). In wild type anaphase I, bivalents separate and 5 pairs of intact univalents segregate to opposite poles, and in anaphase II sister chromatids segregate to form four groups of 5 chromosomes. Mutant meiocytes, by comparison, have fragmented chromosomes in 94% of anaphase I meiocytes, and 93% of anaphase II meiocytes yield abnormal polyads with unequal amounts of DNA ( Figure 1E). These results suggest that the mutant has meiotic DSB repair defects.
To identify the mutated gene, we constructed an F2 mapping population for map-based cloning. We mapped the mutation to a 370 kb interval on chromosome 3 which includes the ATM locus (Figure 2A). The fertility and chromosome fragmentation phenotypes of our mutant are consistent with the known meiotic phenotypes previously described for atm (Garcia et al., 2003). After cloning and sequencing the entire ATM genomic region in our mutant, we found a 766 bp insertion in the intronexon junction of the 49th exon ( Figure 2B and Supplementary Figure S1A). We used a complementation test to validate the identity of our mutant as ATM. We crossed a heterozygous line 184 plant with the previously characterized atm-2 T-DNA allele (Garcia et al., 2003) to generate F1 progeny. In the F1 population, we observed a phenotypic segregation of ten fertile and eight sterile plants. The ATM genotypes correlated with their phenotypes. Cytological analysis further revealed that line 184/atm-2 plants also exhibit chromosome fragmentation phenotypes during meiosis (Supplementary Figure S1C). Taken together, we conclude that the mutated gene in the line 184 is ATM. Following the naming convention of the prior alleles (Garcia et al., 2003;Waterworth et al., 2007;Inagaki et al., 2009), we call our allele atm-5.
To investigate how the insertion affects ATM expression, we designed six pairs of primers corresponding to distinct regions of the coding sequence to detect transcripts by RT-RCR in both wild type and atm-5 ( Figure 2B and Supplementary Table S1). All regions had detectable transcripts in mutant and wild type plants, suggesting the ATM mRNA is not truncated in atm-5. However, the band corresponding to RT4, which includes the insertion site was larger in atm-5 than in wild type ( Figure 2C). Cloning and sequencing ATM cDNA from atm-5 and wild type revealed that the insertion was integrated into the ATM mRNA in atm-5 (Supplementary Figure S2A), potentially due to abnormal splicing. The integrated sequence generates a premature stop codon in the atm-5 mRNA, leading to a truncated protein (Supplementary Figure S2B) lacking its kinase domain (FAT, PI3/4K, and FATC).
In Arabidopsis mitotic cells, a second phosphatidylinositol 3-kinase like family member, Ataxia Telangiectasia and Rad3related protein (ATR), is also involved in the formation of γH2AX (Friesner et al., 2005). To test whether the γH2AX foci in atm meiocytes are dependent on ATR, we examined γH2AX foci in atm-5 atr-2 double mutant meiocytes. The atm-5 atr-2 double mutant is completely sterile due to severe meiotic DSB repair defects, which is consistent with the previously reported phenotype of the atm-1 atr-3 double mutant (Culligan and Britt, 2008). Our data showed that γH2AX foci are nearly abolished (20 ± 15.66, n = 65, P = 4.31E-11) in atm-5 atr-2 early prophase I meiocytes (Figures 3A-D), indicating that the γH2AX foci remaining in atm mutants are ATR-dependent. In atr-2 single mutant meiocytes, the number of γH2AX foci in zygotene (208.48 ± 24.83, n = 33, P = 4.36E-01) and pachytene (72.44 ± 15.66, n = 32, P = 1.86E-02) (Figures 3A-D) are similar with those in wild type. Taken together, these data suggest that ATM is the primary phosphatidylinositol 3-kinase like family member responsible for meiotic γH2AX formation, whereas ATR acts as a back-up when ATM function is compromised.

There Are More Meiotic COs in atm Mutants
To investigate whether ATM affects meiotic recombination, we compared the number of COs in atm and wild type meiocytes using cytological approach. We counted chiasmata, the cytological structures resulting from meiotic COs (Sanchez Moran et al., 2001), in metaphase I cells containing five bivalents from both wild type and atm mutants. The average number of chiasmata per meiocyte in wild type is 10.85 ± 1.18 (n = 20), which is in agreement with previous results (Osman et al., 2011). In the atm mutants the average number of chiasmata is significantly elevated (atm-2: 12.00 ± 1.02, n = 25, P = 2.72E-03; atm-5: 12.01 ± 1.37, n = 21, P = 7.76E-03) (Figures 5A,B and Supplementary Table S6).
randomly without any detectable chromosome fragmentations at anaphase I (Figure 6D5), indicating that meiotic DSBs have been repaired. However, in the double mutant, numerous chromosome fragments are present at anaphase I (Figure 6D6). At anaphase II, the chromosome fragmentation in atm-5 dmc1-3 becomes more obvious after the separation of sister chromatids ( Figure 6E6). Quantification of the number of chromosome fragments shows that the atm-5 dmc1-3 double mutant has a more severe phenotype than the atm-5 single mutant ( Figure 6D2), but is similar to that of rad51-3 For each allele, the anaphase II and telophase II cells were observed after DAPI stained chromosome spreads and centromeric FISH. DAPI stained bodies without centromere signal were taken as chromosome fragments. Chromosome fragmentation levels are divided into four categories: 0 (no chromosome fragmentation), 1-5 (1-5 fragments per cell), 6-10 (6-10 fragments per cell) and >10 (more than 10 fragments per cell). For each allele, the counted cells are from about 12 plants.
and atm-5 rad51-3 ( Figure 7B). Taken together, the additive chromosome fragmentation phenotypes of atm and dmc1 suggest that they act in different meiotic DNA repair pathways in Arabidopsis.

The atm asy1 Double Mutant Has More Chromosome Fragmentation Than Either Single Mutant
ASY1 is thought to be essential for DMC1-dependent inter-homolog recombination during meiosis. Loss of ASY1 eliminates the majority of inter-homolog recombination, resulting in repair of meiotic DSBs using sister chromatids as templates (Sanchez-Moran et al., 2007). We compared meiotic chromosome conformation phenotypes in an atm-5 asy1 double mutant to wild type, atm-5, and asy1. Prophase I meiocytes from asy1 (n = 74) and atm-5 asy1 (n = 102) lack the typical pachytene chromosomes seen in wild type and atm-5 (Figures 6B7,B8). At metaphase I, wild type meiocytes have five well aligned bivalents ( Figure 6C1); and most atm-5 meiocytes have four bivalents (Figures 6C2, 7A and Supplementary Figure S8). In asy1, most meiocytes have a mixture of univalents and bivalents in metaphase I (Figure 6C7), which is consistent with previous observations (Caryl et al., 2000). Interestingly, atm-5 asy1 metaphase I meiocytes have univalents and entangled chromosomes (Figure 6C8), which is a more severe phenotype than either atm-5 ( Figure 6C2) or asy1 (Figure 6C7) single mutants. FIGURE 8 | Model of ATM promote RAD51 mediated inter-sister meiotic DSB repair. After meiotic DSBs are formed, it will be processed to generated the 3' ssDNAs (A). With the help of RAD51 and DMC1, some of the ssDNAs will invade homologous chromosomes to seek template for repair (A,B). Those DSBs repaired using homologous chromosomes will generate COs or NCOs while NCOs are in the majority (B,C). Whereas some of the DSBs are repaired through inter-sister recombination which is mediated by RAD51 (B,C). Thus, we proposed that ATM could promote this RAD51-mediated inter-sister repair process and it is possible that ATM regulate this process through phosphorylating a potential protein X (C). Thus, when ATM is absent, this inter-sister repair pathway was impaired and leads to some unrepaired meiotic DSBs which eventually result in the mild chromosome fragmentations observed in atm mutants.
These entanglements may be due to non-homolog interactions in the double mutant. Homologous chromosomes segregate unequally in asy1 at anaphase I, but there is no detectable chromosome fragmentation ( Figure 6D7). However, in the atm-5 asy1 double mutant, numerous chromosome fragments were observed at anaphase I ( Figure 6D8). The atm-5 asy1 chromosome fragmentation phenotype is more obvious in anaphase II when sister chromatids separate ( Figure 6E8). Quantification of the chromosome fragments showed the phenotype is more severe in atm-5 asy1 compared atm-5, but slightly weaker than that seen in rad51-3, atm-5 rad51-3, or atm-5 dmc1-3 ( Figure 7B). Taken together, the additive chromosome fragmentation phenotypes of atm and asy1 suggest that they work in different meiotic DNA repair pathways in Arabidopsis.

ATM Is Potentially Involved in Meiotic DSBs Formation in Arabidopsis
To facilitate meiotic recombination, each meiocyte needs to generate sufficient DSBs at the onset of meiosis. Arabidopsis is thought to have about 200 DSBs per meiocyte (Choi et al., 2018). But the molecular mechanisms that control the number and distribution of meiotic DSBs are largely unknown. In yeast and mouse, ATM negatively regulates meiotic DSB levels (Cooper et al., 2014). Arabidopsis ATM also has meiotic functions but its influence on meiotic DSBs has not been reported prior to our study. We used immunolocalization of γH2AX as a proxy for meiotic DSBs numbers  and showed that foci numbers were decreased in both atm-2 and atm-5 alleles (Figure 3). In mouse, ATM also mediates γH2AX foci formation (Widger et al., 2018). However, in mice ATM specifically contributes to the formation of γH2AX foci during leptotene (Widger et al., 2018). In Arabidopsis, ATM appears to influence γH2AX foci formation from leptotene to pachytene (Supplementary Figure S4A), suggesting differences in ATM's meiotic function in different organisms. The mild decrease of γH2AX foci number in atm mutant and the elimination of γH2AX foci formation in atm atr double mutant (Figure 3) suggested that ATR is responsible for most γH2AX foci formation in atm mutant. Thus, it is intriguing to study how ATM and ATR choose to regulate different γH2AX foci formation during meiosis further. Studies in other species show that H2AX is directly phosphorylated by ATM during meiosis (Barchi et al., 2008;Cheng et al., 2013). Nonetheless, formal proof that Arabidopsis H2AX is a direct substrate of ATM will require more experimentation. In addition, it is possible that the decrease in γH2AX foci number in atm is due to unphosphorylated H2AX around DSBs, and does not reflect the true number of meiotic DSBs. Thus, it is rational to speculate that the repair of some DSBs without γH2AX localization might be affected in the absence of ATM. However, there is still lack of direct evidence that how γH2AX affect DSB repair. In Arabidopsis, γH2AX seems not required for mitotic DSB repair, as shown by the chromosome fusions observed in the atr rad50 double mutant (Amiard et al., 2010. Meanwhile, mouse h2ax mutant displays male infertility due to unable to form sex body, but female h2ax mouse exhibit normal meiosis (Celeste et al., 2002). Thus, we speculate that it is more likely that the mild reduction of γH2AX in Arabidopsis atm mutant may not lead to meiotic DSB repair defects.
During meiosis, RAD51 and DMC1 binding to the processed single stranded overhangs and facilitating the strand invasion step of DSB repair (Kauppi et al., 2011;Pradillo et al., 2014;Widger et al., 2018). Immunolocalization of RAD51 and DMC1 showed that their foci numbers both increase during zygotene in atm mutants (Figure 4). Given that the conserved meiotic functions of ATM (Harper and Elledge, 2007;McKinnon, 2012), it is possible that Arabidopsis atm mutant also has elevated meiotic DSBs like the case in mouse and yeast atm mutants (Lange et al., 2011;Garcia et al., 2015). As a direct proxy of meiotic DSBs, the increased RAD51 and DMC1 foci in atm zygotene cells also implies that the absence of ATM elevates the number of meiotic DSBs in Arabidopsis. Although Arabidopsis atm mutant might have more meiotic DSBs, chromosome morphology analysis ( Figure 1E) and ZYP1 localization analysis (Supplementary Figure S3) suggest that the synapsis is unaffected when ATM is absent. Besides, we did not observe any univalents in diakinesis cells in atm. Therefore, the increased number of COs in atm meiocytes without early separated homologous chromosomes (Figures 5A,B), which were used to measure CO numbers in this study, could properly reflect the meiotic recombination case in atm. Furthermore, our HEI10 immunolocalization analysis verifies the observation in CO number increasement and indicates that the increased COs in atm mutant are Type I COs (Figures 5C,D). This is consistent with the phenotypes observed in mouse (Barchi et al., 2008) rather than yeast atm mutants in which the increased COs are Type II COs (Anderson et al., 2015).

ATM Plays an Important Role in Regulating Inter-Sister Repair of Meiotic DSBs
Meiocytes usually generate many more DSBs than COs. For example, in Arabidopsis, each meiocyte has over 200 DSBs (Choi et al., 2018), but only 9-10 COs (Lu et al., 2012). To maintain genome integrity, the remaining DSBs are thought to be repaired by the synthesis dependent strand annealing (SDSA) pathway to produce NCOs (McMahill et al., 2007). However, because NCO are difficult to detect unless they are accompanied by a gene conversion event, only a few have been documented in Arabidopsis (Lu et al., 2012;Sun et al., 2012;Wijnker et al., 2013). Furthermore, meiotic DSBs are believed to have a bias for using homologous chromosomes as repair templates (Lao and Hunter, 2010), but the strength of this bias has not be well characterized in plants. Therefore, it is possible that in Arabidopsis a significant number of meiotic DSBs might be repaired using sister chromatids as templates. Unfortunately, there is still no efficient way to directly detect sister-chromatid repair in flowering plants.
Here, we employed genetic studies to investigate functional relationships between the protein kinase ATM and other key meiotic factors, including DMC1, RAD51, and ASY1. In dmc1 single mutants, COs are abolished and bivalents fail to form but there are no detectable chromosome fragments in meiosis (Figures 6, 7; Couteau et al., 1999;Sanchez-Moran et al., 2007), indicating that meiotic DSBs are completely repaired via RAD51-mediated IS repair. In contrast, we found that atm dmc1 double mutants have a severe chromosome fragmentation phenotype not seen in atm or dmc1 single mutants (Figures 6E6, 7B), suggesting that in the absence of ATM, inter-sister repair by RAD51 is impaired in the dmc1 background. This is contrary to budding yeast, in which, ATM (Tel1) indirectly suppress RAD51 through a meiosis-specific kinase Mek1 (Carballo et al., 2008), thus its dmc1 mutant exhibit severe meiotic DSB repair defects (Bishop et al., 1992). Considering that the Arabidopsis dmc1 mutant does not have meiotic DSB repair defects and there was no evidence that Arabidopsis have a homolog of yeast Mek1. It is not surprising that Arabidopsis does not adopted a similar RAD51 suppression pathway. Meanwhile, our results suggest that ATM promoted RAD51-mediated IS repair in Arabidopsis. This idea is further supported by the observation that atm asy1 double mutants have more severe meiotic chromosome fragmentation than atm single mutants (Figures 6, 7B). Since ASY1 is required for DMC1-mediated inter-homolog recombination (Sanchez-Moran et al., 2007), the double mutant is left with unrepaired breaks. In contrast, atm rad51 double mutants have indistinguishable chromosome fragmentation levels compared to rad51 single mutants (Figures 6, 7B), suggesting that ATM and RAD51 both function in IS meiotic repair.
We observed entangled chromosomes at metaphase I in both atm dmc1 and atm asy1 double mutants (Figures 6C6,  7A), which may be the result of ectopic non-homologous recombination. This aberrant phenotype is not observed in the corresponding single mutants, further supporting the idea that ATM functions in a parallel meiotic DSB repair pathway compared to DMC1 and ASY1. There are more RAD51 and DMC1 foci in atm meiocytes, but the increase in zygotenestage RAD51 foci in atm meiocytes is much larger than that of DMC1 foci, indicating that ATM may affect the balance of RAD51 and DMC1 loading onto repair intermediates during meiosis. Moreover, we observed that atm pachytene meiocytes accumulate more RAD51 foci than that wild type, while DMC1 pachytene foci are at wild type levels (Figure 4). This suggests that RAD51-mediated DSB repair is delayed or impaired in atm mutants. Thus, we speculate that the DSB repair defects observed in atm mutants are due to impairment of the IS repair pathway.
In summary, we propose a working model for ATM in meiotic recombination. In the model, after meiotic DSBs formation and resection (Figure 8A), the single strand DNA tails are coated with RAD51 and DMC1 and seek DNA templates to facilitate repair ( Figure 8B). In wild type, those DSBs are repaired primarily by inter-homolog interactions, generating meiotic COs or NCOs ( Figure 8C). However, even in wild type, some DSBs may be processed by RAD51-mediated intersister repair with the assistance of ATM (Figures 8B,C). In the absence of ATM, inter-homolog repair is still be mediated by DMC1 but IS repair is significantly compromised (Figure 8C), leading to unrepaired breaks and chromosome fragmentation. As ATM is an evolutionary and functionally conserved kinase (Xu et al., 1996;Garcia et al., 2003;Jones et al., 2012), its role in regulating IS repair may have relevance for other species including animals and fungi.

Plant Materials and Growth Conditions
The atm-2 mutant used in this study were described in previous studies (Garcia et al., 2003;Waterworth et al., 2007). Another atm mutant is the newly identified atm-5. All the following quantitative analysis include these two independent atm mutants. Other mutant alleles used in this study are spo11-1-1 (Grelon et al., 2001), rad51-3 ), asy1 (Sanchez-Moran et al., 2007, hei10-2 (Chelysheva et al., 2012), and atr-2 (Culligan et al., 2004). The dmc1-3 allele was a previously identified Ds insertion mutant in our laboratory. The Ds insertion is located in the first intron of DMC1 (Supplementary Figure S6A). This expression of DMC1 in dmc1-3 is undetectable using primers across the insertion site (Supplementary Figure S6B). The atm-5 atr-2, atm-5 dmc1-3, and atm-5 rad51-3 double mutants were generated by crossing the corresponding heterozygous mutants with heterozygous ATM/atm-5 plants. All materials were grown at 20-22 • C under long day conditions (16 h light/8 h dark). Primers used for genotyping are listed in Supplementary Table S1.

Characterization of Mutant Phenotypes
Pollen viability was analyzed through Alexander staining on predehiscent anthers according to a simplified protocol (Peterson et al., 2010). For tetrad morphology analysis, tetrad-stage male meiocytes were squeezed from fresh anthers onto glass microscope slides and stained with Toluidine Blue solution (0.025%, m/v). After covering with cover glass, the stained tetrads were examined using an Axio Imager A2 microscope (Zeiss, Heidelberg, Germany) under white light illumination.

Map-Based Cloning and Mutation Identification
To construct the mapping population, we crossed the heterozygous atm-5 (Ler background) to wild type Col-0. An F2 population was generated by allowing F1 plants to self-fertilize. From 586 F2 plants, 143 sterile plants were isolated and used for mapping. Genetic markers for map-based cloning of the atm-5 mutation were designed based on simple sequence length polymorphisms (SSLP) between Columbia (Col-0) and Landsberg erecta (Ler). Primers for amplifying SSLPs are listed in Supplementary Table S2. The entire ATM gene was divided to seven overlapping segments for cloning by PCR. Each segment was cloned into the pEASY-T1 vector (CT111-02, Transgen Biotech, Beijing, China) and a single colony containing for each segment was selected for sequencing. The full cDNA of ATM was also cloned using a similar approach with six segments into the cloning vector for sequencing. Primers sequences are listed in Supplementary Table S1.

Expression Analysis
Semi-quantitative RT-PCR was used to analyze the expression of ATM in different atm mutant alleles. Total RNA from wild type and mutant inflorescences was isolated using the Trizol reagent (Cat # : 15596018, Thermo Fisher Scientific). The first strand of cDNA was synthesized using the PrimeScript 1st strand cDNA synthesis kit (Cat # : 6110B, Takara, Beijing, China). Semiquantitative RT-PCR was conducted with the 2 × Taq master mix (Cat # : E005-01B, Novoprotein, Shanghai, China). Primers sequences are listed in Supplementary Table S1.

Chromosome Spread and Immunolocalization Analyses
Chromosome spreads and centromere FISH were conducted as described previously (Wang et al., 2014). Immunolocalization experiments were performed following the previous description (Armstrong et al., 2009) with minor modifications. Briefly, after chromosome spreads were prepared, the slides were washed in wash buffer I (PBS, 1% Triton X-100, pH 7.4) for 1 h. Then these slides were blocked at 37 • C for 1 h with blocking buffer (PBS, 5% BSA, 1 mM EDTA, pH 7.4). The primary antibody was diluted to working concentration with blocking buffer and then added to the slides. The slides were covered with parafilm and incubated overnight at 4 • C in a moisture chamber. After incubation, these slides were washed with washing buffer II (PBS, 0.1% Tween 20, pH 7.4) three times (15 min each time). The secondary antibody was diluted to working concentration with blocking buffer and then added to the slides. After covering with parafilm, the slides were incubated at 37 • C in a moisture chamber in dark. The slides were then washed with washing buffer II three times (15 min each time) and mounted in 1.5 µg/mL DAPI (Vector Laboratories, Burlingame, CA, United States). The working concentration of primary antibodies are γH2AX at 1/200 dilution, DMC1 at 1/800 dilution, RAD51 at 1/50 dilution, ZYP1 at 1/50 dilution and ASY1 at 1/200 dilution. Immunolocalization of SYN1 (1/200 dilution), HEI10 (1/400 dilution) and ASY1 (1/200 dilution) was carried out as described previously (Chelysheva et al., 2013). Rabbit polyclonal RAD51, ASY1, SYN1 and rat polyclonal ZYP1 antibodies were generated and verified previously (Wang et al., , 2016Niu et al., 2015). Rabbit polyclonal DMC1 antibody was generated as described previously (Wang et al., 2019). Rabbit polyclonal γH2AX antibody was generated and tested previously (Zhou et al., 2016). The HEI10 polyclonal antibody was generated through immune rabbit with a KLH conjugated peptide sequence of HEI10 (PKDEIWPARQNS). The γH2AX, HEI10 and DMC1 antibody was further verified through immunolocalization in spo11-1-1, dmc1-3, hei10-2 respectively (Supplementary Figure S7). Secondary antibodies used in this study were Alexa Fluor 488 Goat Anti-Mouse IgG (H + L) (A32723, 1/1000 dilution), Alexa Fluor 555 Goat Anti-Rabbit IgG (H + L) (A32732, 1/1000 dilution) and Alexa Fluor 488 Goat Anti-Rabbit IgG (H + L) (A32731, 1/1000 dilution) (Invitrogen, Carlabad, CA, United States). Microscopy was carried out using an Axio Imager A2 microscope (Zeiss, Heidelberg, Germany). Images were processed with Image J (Collins, 2007) and figures were constructed using Photoshop CS3 and Illustrator CS6 (Adobe, Mountain View, CA, United States). To count the number of fluorescent foci signals, images of chromosome signal and protein foci signal were merged together without any processing. Only those foci signals merged onto chromosome signal were counted through the Image J count tool. The cells used for counting γH2AX foci, RAD51 foci, DMC1 foci or HEI10 foci in each allele are from at least 6 slides. Each slide contains young flower buds from about 5 plants. The cells used for chiasmata counting are from about 15 slides and each slide contains samples from two plants. The cells used for bivalent counting are from about 10 slides and each slide contain samples from two plants. The cells used for chromosome fragments counting are from about 6 slides and each slide contains samples from two plants. The quantitative analysis of γH2AX, RAD51, DMC1 foci, HEI10 foci number and chiasmata number in atm mutant both include two independent mutant alleles (atm-2 and atm-5). The qualitative analysis of γH2AX foci localization and meiotic chromosome morphology all include one representative single or double mutant allele.

Statistical Methods
Statistical comparisons of the γH2AX foci, RAD51 foci, DMC1 foci, HEI10 foci and chiasmata number was carried out in R using the Wilcoxon Rank Sum test.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.

AUTHOR CONTRIBUTIONS
PL and YY designed the research. YY, WC, HL, LM, DR, and AM performed the genetic mapping experiments and molecular analysis. YY performed the molecular and cytological analyses. YY, XL, and PL analyzed the data. YY wrote a draft for the manuscript. PL revised the draft and wrote the manuscript. All authors read and approved the final manuscript.

FUNDING
This work was supported by a grant from the National Natural Science Foundation of China (No. 31770351 to PL) and start-up funding from Henan University to PL. PL also appreciate the great support from Rijk Zwaan for meiosis studies in his lab. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2020.00839/ full#supplementary-material Blast results of this 766 bp segment against the Arabidopsis genome sequence. In each chromosome, there exist a locus that is highly similar with this segment. (C) Male meiosis chromosome morphology at anaphase I of atm-2/atm-2 and line 184/atm-2 (n = 2). Bar = 5 µm. FIGURE S3 | Immunolocalization of AYS1, ZYP1 and SYN1 in male meiocytes of atm mutants and wild type. (A) Zygotene stage chromosomes (blue) with ASY1 signal (red) and ZYP1 signal (green). (B) Pachytene stage chromosomes (blue) with ASY1 signal (red) and ZYP1 signal (green). (C) Chromosomes with SYN1 (red) signal in wild type and atm mutants at different meiotic stages. Chromosomes were stained with DAPI (blue). In each line, at least 20 meiocytes were examined. Bar = 5 µm. The atm-2 and atm-5 are two independent atm mutant alleles. FIGURE S4 | γH2AX localization in leptotene and pachytene meiocytes of wild type and atm mutants. (A) Immunolocalization of γH2AX in leptotene meiocytes of wild type and atm mutants. (B) Dual-immunolocalization of γH2AX (red) and ZYP1 (green) in pachytene meiocytes of wild type and atm mutants. Chromosomes were stained with DAPI (blue). Bar = 5µm. The atm-2 and atm-5 are two independent atm mutant alleles.