Dynamic localization of SPO11-1 and conformational changes of meiotic axial elements during recombination initiation of maize meiosis

Meiotic double-strand breaks (DSBs) are generated by the evolutionarily conserved SPO11 complex in the context of chromatin loops that are organized along axial elements (AEs) of chromosomes. However, how DSBs are formed with respect to chromosome axes and the SPO11 complex remains unclear in plants. Here, we confirm that DSB and bivalent formation are defective in maize spo11-1 mutants. Super-resolution microscopy demonstrates dynamic localization of SPO11-1 during recombination initiation, with variable numbers of SPO11-1 foci being distributed in nuclei but similar numbers of SPO11-1 foci being found on AEs. Notably, cytological analysis of spo11-1 meiocytes revealed an aberrant AE structure. At leptotene, AEs of wild-type and spo11-1 meiocytes were similarly curly and discontinuous. However, during early zygotene, wild-type AEs become uniform and exhibit shortened axes, whereas the elongated and curly AEs persisted in spo11-1 mutants, suggesting that loss of SPO11-1 compromised AE structural maturation. Our results reveal an interesting relationship between SPO11-1 loading onto AEs and the conformational remodeling of AEs during recombination initiation.


Introduction
Meiosis is an essential cell division that halves ploidy numbers through two successive chromosome divisions to produce gametes for sexual reproduction. The faithful segregation of homologous chromosomes requires meiotic recombination during prophase I, which mediates pairing of homologous chromosomes and creates physical connections until their segregation [1]. Meiotic recombination is initiated by the introduction of double-strand breaks (DSBs) into chromosomal DNA via SPO11, a protein of the topoisomerase VI family [2]. The SPO11 complex cuts double-stranded DNA and links covalently to the 5' ends at the break point [3]. Following SPO11 removal together with 5'-oligonucleotides, DSBs are further resected to produce 3' single-stranded DNA (ssDNA) tails for searching homologues with the assistance of RAD51 and DMC1 recombinase proteins. Only a small subset of DSBs are repaired using a chromatid of homologous chromosomes to give rise to crossovers (COs), whereas most DSBs lead to non-crossovers (NCOs) [4]. These CO sites, seen as chiasmata at the diplotene stage, ensure that pairs of homologs, i.e. the bivalents, maintain their association at the metaphase plate, which allows subsequent reductional segregation at anaphase I. Thus, at least one CO per bivalent is needed for accurate segregation of homologous chromosomes. When DSB formation is defective, such as in spo11 mutants, meiotic recombination is impaired in all species analyzed to date, leading to infertile or aneuploid gametes [5][6][7][8][9][10].
These essential meiotic DSBs are also potentially harmful to cells, so their formation is tightly regulated both temporally and spatially to allow proper initiation of recombination [1]. For example, the timing of DSB induction is linked to the completion of pre-meiotic DNA replication [11] and it is restricted to a narrow timeframe [12]. Moreover, the total number of DSBs is also regulated, as too few DSBs could be detrimental given that sufficient DSBs are required for homologous engagement in most species. However, too many DSBs can also be deleterious since they may result in chromosome fragmentation if unrepaired. Furthermore, clustered DSBs in a short interval is also likely prevented [13]. Recent studies in yeast and mouse have revealed that DNA damage-responsive kinase Tel1/ATM activity and subsequent homologous engagement are involved in negative feedback loops to restrain SPO11 from marking too many DSBs [13][14][15]. However, the control mechanism and its regulatory network for initiating and then halting DSB formation are not yet well understood.

Maize mutants mtm99-14 and mtm00-03 exhibit meiotic phenotypes
We used a forward genetic approach with active Mutator (Mu) transposition [47,48] and identified mtm99-14 and mtm00-03 which are defective in male and female fertility (S1 and S2 Tables). In maize, male meiosis occurs within anthers that develop progressively with respect to their positions on a male tassel. Meiotic progression, especially for prophase I, is correlated nicely with anther length, providing a reliable means of staging and comparing wild-type (WT) and mutant anthers [49]. The Fig 1A displays that DAPI-stained chromosomes of leptotene meiocytes appear as long, thin threads with multiple spheres of heterochromatic knobs, and that a nucleolus is located within the nuclear interior. A profound chromosomal reorganization takes place in maize meiocytes at late leptotene, seen as that chromatin mass is polarized against an offset nucleolus and heterochromatic knobs exhibit transient elongation [50]. During WT zygotene, homologous chromosomes pair gradually, often manifesting as thicker chromosomes and cylinder-shaped heterochromatic knobs in synapsed regions. At pachytene, synapsis is complete with chromosomes exhibiting a uniform appearance along their lengths except for the heavily-stained rounded heterochromatic knobs. Meiosis in the mtm00-03 and mtm99-14 mutants appeared to progress normally until the zygotene stage, with an obviously abnormal chromosome phenotype in both mutants at pachytene stage (Fig 1A).
At the metaphase I, ten bivalents align on the metaphase plate in WT, but mtm99-14 and mtm00-03 meiocytes mainly contained univalents with occasional bivalents. In contrast to normal WT tetrad cells, mutants exhibited abnormal microspores as unequal tetrads, tetrads with micronuclei, or polyads (Fig 1B and 1C). The reduced fertility of mutant maize ears suggested that female meiosis is also affected. Only a few seeds were obtained from mutants when they were pollinated with WT pollen (S2 Table and S1 Fig). The proportion of seeds obtained from mutants is not statistically different to the expected proportion arising from completely random segregation of ten pairs of chromosomes (ε = 0.82 for mtm99-14, ε = 0.32 for mtm00-03, < ε α5% = 1.96). Given the similar phenotypes observed, we performed an allelism test and confirmed that the mtm99-14 and mtm00-03 mutants are allelic (Fig 1B and 1C and S1 and S2 Tables).

Identification of mutations in the Spo11-1 locus
We mapped the mtm00-03 mutation to chromosome 5 by molecular markers (see Material and Methods). To identify the gene, Mu transposon insertions that are associated with the sterile phenotype were isolated using the Mu-Illumina approach [51]. We found that the mtm99-14 and mtm00-03 mutants harbor traces of Mu sequences associated with deletions spanning the Spo11-1 (GRMZM2G129913) and Phytochrome C2 (PhyC2, GRMZM2G129889) genes ( Fig  2A). We could not determine the lengths of those deletions, and only one Mu border was recovered in our Mu-Illumina sequence analyses. Nevertheless, reverse transcription PCR experiments confirmed that neither the mtm99-14 nor mtm00-03 mutant expressed the Spo11-1 gene in meiotic anthers (Fig 2B). Although the PhyC2 gene is deleted, the closely paralogous gene PhyC1 (GRMZM2G057935) that shares 94% identity with PhyC2 is expressed in both mutants and it may be sufficient for phytochrome C function (S2 Fig).
Considering complex rearrangements that may occur in both the mtm99-14 and mtm00-03 mutants, we searched for other mutations affecting only the Spo11-1 gene. From the Trait Utility System for Corn (TUSC) transposon mutagenesis population [52], we identified one additional mutant, named spo11-1-1, containing a Mu insertion in the first exon (Fig 2A). We showed that the spo11-1-1 mutant exhibited identical phenotypes to the mtm99-14 and mtm00-03 mutants, and an allelism test confirmed that the spo11-1-1 mutant is allelic (S3 Fig  and S1 Table). Accordingly, we renamed mtm99-14 and mtm00-03 mutants as spo11-1-Δ14 and spo11-1-Δ03, respectively. Since all alleles exhibited an identical phenotype, we characterized the spo11-1-Δ03 mutant in a faster growing background extensively and included some results of other alleles in supplemental figures (S3, S20 and S21 Figs).

Maize Spo11-1 exhibit alternative splicing
Our RT-PCR analysis of Spo11-1 expression detected two spliced variants, which arose from retention of the second intron (Fig 2B and 2D). Both transcripts are predicted to produce Projected images of WT and mutant meiocytes stained with DAPI during early prophase I. Initially spherical heterochromatic knobs (arrows) become elongated at the leptotene-zygotene transition that accompanies a positional change of the nucleolus (n) from the nuclear interior to the periphery. At zygotene, paired knobs (arrowhead) and chromosomes (dashed line) were observed in WT. An obvious aberrant phenotype of both mutants was observed during the pachytene stage, manifested as fluffy chromosomes. Scale bar represents 5 μm. (B) Acetocarmine staining of male meiocytes during the metaphase I (upper panel) and tetrad stages (lower panel) of WT, mtm00-03, mtm99-14, and heteroallelic mtm99-14/mtm00-03 plants. In WT male meiocytes, ten bivalents are aligned on a metaphase I plate and later produce normal tetrad cells. Mutant cells mainly present univalents and rare interconnected chromosomes counted as bivalents (arrows). Scale bar represents 10 μm. (C) Quantitative results of meiotic phenotypes during the metaphase I and tetrad stages. Numbers of cells analyzed are shown in parentheses. The spo11-1-1 is another mutant allele (see next section).
https://doi.org/10.1371/journal.pgen.1007881.g001 functional SPO11-1 proteins because they contain a nuclear localization signal (KLRR), the topoisomerase-primase 'TOPRIM' domain that is essential for catalytic activity, and the conserved tyrosine residue in the winged-helix DNA-binding (WHD) domain ( Fig 2E). The larger protein isoform (named Spo11-1β) contains an extra 43 amino-acid domain, which is predicted to form an alpha-helix motif upstream of the TopVIA type endonuclease domain (S4 Both mtm99-14 and mtm00-03 mutants contain remnants of Mu sequences (purple boxes) that are associated with deletions (grey boxes) spanning PhyC2 and Spo11-1 genes. Coding regions and introns/UTRs are represented as orange and yellow boxes, respectively. The additional Mu insertion allele (spo11-1-1) is indicated by a red triangle. Green boxes represent transposable elements (Tourist and CACTA-like elements). (B) RT-PCR analyses to amplify the Spo11-1 at 5' and 3' regions exhibited splicing variants (asterisks) in WT and confirmed that Spo11-1 transcripts are absent in all mutant alleles. Positions of primer sets are indicated by red arrows in (D). Afd1/Rec8 expression was used as a control. 0RT represents a negative control lacking cDNA. (C) Quantitative PCR showed differential expression of the Spo11-1α and Spo11-1β transcripts using specific primers indicated by black arrows in (D). (D, E) Gene models (D) and predicted proteins (E) of SPO11-1α and SPO11-1β are shown. The additional region of SPO11-1β (amino acid 73-115) is indicated by a pink box.
https://doi.org/10.1371/journal.pgen.1007881.g002 and S5 Figs). We designed transcript-specific primers (S7 Table) to examine temporal expression of the two transcript variants. Quantitative PCR results revealed that Spo11-1α is more strongly expressed during leptotene, whereas the Spo11-1β transcript level is higher than the Spo11-1α transcript in early zygotene anthers ( Fig 2C). Unfortunately, our attempts to establish if these two isoforms have different functions were not successful.
That DNA damages, possibly including DSBs, were detected in spo11-1 mutants, albeit at very low levels, was unexpected. Consequently, we sought to assess homologous pairing in the absence of the Spo11-1 gene by fluorescence in situ hybridization (FISH). We found that 51% of WT meiocytes at zygotene (n = 15/29) and all pachytene meiocytes (n = 15) displayed paired 5S rDNA signals (Fig 3E and 3F). In contrast, among 28 zygotene-like and 31 pachytene-like spo11-1 meiocytes, we did not detect any paired 5S rDNA signals, even though telomere clustering appeared to be normal ( Fig 3G). During diakinesis and metaphase I, all WT cells (n = 46) exhibited nicely paired 5S rDNA signals, whereas all spo11-1 meiocytes (n = 41) showed separate 5S rDNA signals (Fig 3I-3L). Among mutant cells, we observed a bivalent aligned at the metaphase plate with a single 5S rDNA signal, and the other 5S rDNA signal emanated from a separating chromosome (Fig 3L), suggesting that the occasional bivalents observed in spo11-1 mutants were likely derived from non-homologous association. Taken together, these results suggest that normal DSB formation, chromosome pairing and homologous recombination is abolished in spo11-1 mutants.

Detection of numerous SPO11-1 foci in nuclei
To better understand SPO11-1 localization, we generated a SPO11-1 antibody against 28 amino acids at the N-terminal region and validated its specificity using peptides in the corresponding regions of SPO11-1, SPO11-2 and SPO11-3 ( Fig 4C and S7 Fig). WT meiocytes from early leptotene to pachytene were subjected to immunolocalization and observed using Delta-Vision deconvolution microscope (Fig 4A). At early leptotene, most SPO11-1 signals were detected around the nuclear periphery. We observed approximately 2000-3000 SPO11-1 foci in WT nuclei during late leptotene, when the chromosome mass becomes polarized. Localization of SPO11-1 in nuclei remained until the pachytene stage ( Fig 4D and S3 Table). In contrast, our SPO11-1 antibody detected significantly fewer foci in spo11-1-Δ03 meiocytes ( Fig  4B), similar to background levels in WT meiocytes using pre-immune serum (S7 Fig). To further study the dynamic localizations of SPO11-1 in detail, we took advantage of the slowly progressing and well-defined meiotic stages in maize and preserved the spatial organization of meiocytes in a polyacrylamide gel for subsequent immunolocalization. Super-resolution analyses by structured illumination microscopy of snapshots of SPO11-1 distribution Meiocytes at diakinesis and metaphase I from WT and spo11-1-Δ03 hybridized with 5S rDNA (green, arrows) and 180bp-knob probes (white signals) showing that normal crossover formation is revoked in spo11-1 meiocytes. Note that a bivalent detected in Fig 3L represents   relative to axial protein DSY2 enabled a better depiction of DSB formation with respect to chromosome axes. In addition, the appearance of chromosomes and chromosome axes as well as the lengths of anther samples and their relative positions on a tassel provided an accurate time course for meiotic progression. In Fig 5, we show representative images of nuclei displaying SPO11-1 signals and DSY2-labeled chromosome axes from the leptotene to zygotene stages (respective montages of single Z images are shown in S8-S11 Figs). In agreement with our deconvolution microscopy results (Fig 4), SPO11-1 signals were first observed mostly around the nuclear periphery and with fewer foci within nuclei (Fig 5A and 5B) and were then distributed inside nuclei when the nucleolus become offset at late leptotene (Fig 5D and 5E). When visualized by single Z sections (Fig 5B, 5C, 5E, 5F, 5H, 5I, 5K and 5L), SPO11-1 signals were observed as punctate foci, some of which were colocalized with DSY2-labeled AEs. To count numbers of foci, we extracted SPO11-1 signals in nuclei using a 3D Spot Segmentation algorithm and conducted quantification in the 3D ROI Manager of ImageJ [53] (S12 Fig). We carefully analyzed a total of eighteen meiocytes, and again detected thousands of SPO11-1 foci distributed in nuclei ( Fig 6J and S4 Table). Interestingly, we recorded a wide range of SPO11-1 foci per meiocyte at each meiotic stage. In contrast, using the same detection conditions and image processing method, a significantly lower number of SPO11-1 foci was observed in spo11-1-Δ03 mutants (S13 Fig and S4 Table).

Similar numbers of SPO11-1 foci load onto axial elements
To analyze SPO11-1 distributions relative to chromosome axes in 3D space, we developed an image processing pipeline using ImageJ [54]. The method uses the Trainable Weka Segmentation plugin, a machine learning algorithm, to segment chromosome axes (S14 Fig) [55]. The Fig 6 presents DSY2 and SPO11-1 signals in a representative zygotene meiocyte that was subjected to our image processing pipelines (S12 and S14 Figs). After segmentation, we examined SPO11-1 objects ( Fig 6E) and skeletonized DSY2-labeled chromosome axes ( Fig 6F) by merging the respective images with original images and visualized them using 3D Viewer tool in ImageJ. As shown in a surface-rendered rotation movie ( Fig 6I and S1 Video), most SPO11-1 foci are distributed in between chromosome axes likely filled with DNA, whereas only a few SPO11-1 foci are found on chromosome axes.
To measure the physical distances between SPO11-1 foci and the nearest chromosome axis, we computed the skeletonized axis model representing medial lines along chromosome axes in 3D stacks and generated a 3D distance map which assigns propagating distances from each point along axes to neighboring voxels in the 3D space (S15 Fig). By applying segmented SPO11-1 objects onto the corresponding 3D distance map, we obtained the geodetic distance between each SPO11-1 object and the closest DSY2-labeled axis (S16 Fig and S5 Table). In the S17 Fig, we display the distribution of these minimal distances in each meiocyte and show that a majority of SPO11-1 objects was located within a radius of 0.6 μm from a neighboring axis. More than 50% of the foci in each meiocyte were distributed within DAPI-stained chromatin regions (a radius of 0.3-0.4 μm) in most of the cells examined. We defined SPO11-1 signals as AE-associated objects when the minimal distance to the closest chromosome axis was zero (S5 Table). Surprisingly, despite a wide range (~1691-4750) of SPO11-1 foci being distributed in nuclei, approximately 300-400 SPO11-1 foci were associated with chromosome axes during recombination initiation and those numbers decreased slightly during zygotene ( Fig 6J and S4 Table).

Maize spo11-1 mutants present an aberrant axial element structure and promiscuous synapsis
To better characterize phenotypes of chromosome axes in the absence of Spo11-1, we compared AEs of WT and mutant meiocytes sampled from similarly sized anthers by DSY2 immunostaining. Our results show that WT and mutant meiocytes at leptotene possessed long and curly AEs enclosing a large, centered nucleolus (Fig 7A and 7B). At late leptotene, the chromosome reorganization that offsets the nucleolus position occurs normally in spo11-1 mutants (Fig 7C and 7D). In WT, pairing and synapsis take place mostly from telomere clustering regions and extend toward interstitial regions, often manifesting as multiple aligned bands of bright DSY2 signal intensity ( Fig 7E). However, in contrast to WT meiocytes in which AEs became smoother and more linear, we observed curly and aberrant AEs with uneven DSY2 accumulation in spo11-1 mutants (compare Fig

PLOS GENETICS
and 7H). The erratic structure of these AEs become more evident when we visualized the projection images of a few Z-sections (see magnified views in Fig 7A-7H and S18 Fig). To describe and quantify the cytological aberrance of these AEs, we measured the total lengths of chromosome axes (S14 Fig) and also computed their curvatures [56] (S19 Fig). In Fig 7I, we show that AE length diminished as meiosis progressed in both WT and mutant meiocytes (S6 Table). Notably, AEs in WT meiocytes shortened from~4 to~3 mm during the leptotene to late leptotene stages before synapsis took place, but a corresponding curtailment was not found in mutant meiocytes. Moreover, kappa values of AE curvature in mutant cells were elevated in late leptotene and zygotene stages relative to WT (Fig 7J), suggesting that not only is compactness of AEs affected, but their morphology is also altered, appeared as more flexuous axes in spo11-1 mutants.
Super-resolution images of ASY1 and DSY2 staining in WT meiocytes revealed diminished ASY1 intensity and enhanced DSY2 intensity in synapsed regions, which reflects the progression of synapsis ( Fig 8A). In mutant meiocytes, DSY2 and ASY1 both exhibited a similar abnormality (Fig 8B), seemingly due to incorrect patterning of axial proteins. Similarly, curly and unevenly accumulated ASY1 staining were observed in spo11-1 mutants (Fig 8C and 8D). Despite the aberrant AEs and the lack of DSBs, all maize spo11-1 mutants showed delayed synapsis as short stretches of ZYP1 signal appear later than those observed in WT (Fig 9C and S20  Fig). In contrast to WT at pachytene, mutants displayed discontinuous ZYP1 bands in tangled chromosomes, likely between non-homologous chromosomes (Fig 9D). Transmission electron microscopy of silver nitrate stained SCs revealed some synapsed regions with fold-back conformations, suggesting promiscuous synapsis (Fig 9G and S21 Fig). Together with our FISH results (Fig 3H and 3L), we conclude that synapsis is initiated randomly between nonhomologous chromosomes in the spo11-1 mutants.

Discussion
Molecular and genetic studies over past decades have garnered substantial information on meiotic DSB formation. First, programmed DSBs in meiosis require SPO11 and the meiotic topoisomerase VI subunit B protein, forming a complex [57], as well as other DSB-accessory proteins [40]. Second, DSBs form preferentially in open chromatin regions residing within chromatin loops that emanate from chromosome axes [58]. Third, potential meiotic DSB sites are brought to chromosome axes [40,59,60]. Fourth, the SPO11-containing complex is activated and then generates a DSB during or after the loop-tethering event [1]. These findings favor the 'tethered loop-axis complex' model and also explain the observation of a hierarchical combination of factors governing DSB sites. Although much is known about this process, a full picture of is still elusive.
One aspect that remains unclear is how DSBs form in relation to SPO11 localization and chromosome axes. It is difficult to detect SPO11 in most organisms, probably due to its narrow timeframe of expression, its dynamic localization, and/or low sensitivity for in situ detection of its protein molecules. It has been observed by immunocytochemistry of chromosome spread Only linear contrast adjustment was applied to enhance signals. Scale bars represent 5 μm. (D-F) Projection images of segmented SPO11-1 objects (E) and a skeletonized axis model (F) representing the medial lines along chromosome axes generated using our image processing pipelines (S12 and S14 Figs). (G-H) Surface rendering images of SPO11 and DSY2, visualized by the 3D Viewer tool in ImageJ. Scale bars represent 5 μm. (I) A magnified 3D view of the surface rendering image in G. Scale bars represent 1 μm. A respective rotation movie is presented in S1 Video. (J) Numbers of total SPO11-1 objects (dots) and axis-associated SPO11-1 objects (triangles) detected in meiocytes during early prophase I.  . (A, B) The morphology of chromosome axes is similar in WT and mutant meiocytes when DSY2 first appears in the nucleus at the early leptotene samples in S. cerevisiae [61], S. macrospora [62], Arabidopsis [63] and mouse [64]. These results show that SPO11 foci appear on chromosome axes and/or in chromatin regions during the leptotene to early pachytene stages. Unfortunately, due to limitation of microscope resolution and slide preparation of chromosome spreads, it seems difficult to analyze SPO11 localization in detail.
However, due to the remarkable advantages of maize for cytological observations, our analyses have led to several interesting observations that provide new insights into DSB formation. By super-resolution microscopy and sophisticated image processing on 3D-preserved maize meiocytes, we have uncovered dynamic SPO11-1 localization during recombination initiation. First, SPO11-1 signals were observed mostly around the nuclear periphery and later distributed in nuclei. Next, although a broad range of thousands of SPO11-1 foci was detected in nuclei with more than half of them colocalized with DAPI-stained chromatin, only relatively similar numbers (~350) of SPO11-1 foci were associated with chromosome axes (Fig 6). It is intriguing that we detected a large amount of SPO11 distributed around or within chromatin regions since the SPO11 enzyme is intrinsically a threat for genome integrity. However, this finding may reflect the fact that SPO11 localization on chromatin loops alone is inadequate to make DSBs. Moreover, previous studies have suggested that 200-500 DSBs are generated during each round of maize meiosis [65], which is in accord with numbers of AE-associated SPO11-1 signals detected in this study, implying that only AE-associated SPO11 foci can generate DSBs. In addition, the timing of SPO11-1 localization on chromosome axes coincides with the leptotene to zygotene transition when homologous alignment and telomere clustering are taking place. Taken together, our results suggest a model whereby numerous SPO11 proteins bind on potential DSB sites, and the competitive and dynamic tethering events that link SPO11-marked hotspots to chromosome axes may play a critical role in DSB formation. It is not clear if the SPO11-1 foci on chromatin represent DSB hotspots in our analyses. However, in support of this notion, a previous study in S. cerevisiae using a standard ChIP protocol has detected a transient and noncovalent association of SPO11 with meiotic hotspots [61]. That transient association is independent of DSB formation since it was also detected in spo11-Y135F mutation, suggesting that SPO11 accumulates transiently at hotspots before stable DSBs form at some of sites. It is also interesting to note that maize SPO11-1 signals remained until the pachytene stage which is consistent with previous finding of detecting SPO11-1-myc foci in the Arabidopsis SPO11-1-myc spo11-1 mutant line [63].
In our study, we have characterized the phenotype of maize spo11-1 mutants in detail and showed that loss of Spo11-1 in maize results in meiotic sterility, as has been observed in other species. Unexpectedly, we also observed abnormal chromosomal axes in maize spo11-1 mutants. First, leptotene AEs in mutants were indistinguishable from those of WT in terms of their length, morphology and curvature. However, differences became apparent when meiosis proceeded to late leptotene when a global re-organization of chromosome configuration takes place. Our results show that spo11-1 meiocytes displayed longer chromosome axes with aberrant and uneven accumulation of DSY2 and ASY1 patches, unlike the smooth and shortened chromosome axes of WT meiocytes. Moreover, at early zygotene when synapsis is initiated stage. (C, D) At late leptotene, AEs appear progressively as linear forms in WT (C). In contrast, abnormal AE structures are observed in mutant meiocytes (D). Note that DSY2 intensity is uneven in mutant meiocytes, with atypically bright fluorescent patches, unlike the uniform signals in the WT. The off-set nucleolus is marked with n. (E-F) Synapsis, manifested as paired doubled DSY2 stretches, is observed primarily near the telomere bouquet ( � ) of the WT (E), which is absent from mutant meiocytes (F).(G-H) A late zygotene WT meiocyte (G) showing consistent DSY2 signal along chromosomes, unlike the unevenly accumulated DSY2 (arrows) in spo11-1 meiocytes (H).(I) Dot plot graph showing total lengths (mm) of DSY2 signals from the leptotene to pachytene stages in WT and spo11-1 meiocytes. Each dot represents a measurement from one meiocyte. The linear plot represents the percentage of synapsed regions, as measured by ZYP1 signal length over total DSY2 signal length.(J) Distribution of kappa values representing AE curvature, as measured from five meiocytes for each stage in WT and spo11-1 mutants. https://doi.org/10.1371/journal.pgen.1007881.g007 shortly after pre-alignment in WT, mutant cells showed delayed SC nucleation and aberrant chromosome axes. The curly AEs in spo11-1 meiocytes may result from mis-patterning of axial element proteins, which appear after the late leptotene stage. This observation points to the existence of a possible mechanism that is established in response to DSBs as a part of AE maturation process. For example, when DSBs are made in recombinosomes that are located on

PLOS GENETICS
chromosome axes, a localized structural transition may be triggered in the chromosome axes such as by remodeling axial proteins [66], by phosphorylation of ASY1/HOP1 and REC114/ PHS1 via the ATM-mediated DNA damage response [15,67], or simply by AE condensation, to prepare the sites for sequential interhomolog recombination or to prevent additional DSB  H). Note that synapsis occurs within one axis, forming "fold-back" structures (arrow), and promiscuous synapsis can also be seen as unequal synapsis with an overhang (arrowheads). In addition, unsynapsed AEs in mutants appear as curly and twisted structures (asterisk in H). Scale bars represent 1 μm.
https://doi.org/10.1371/journal.pgen.1007881.g009 generation by rejecting further tethering events. In the absence of SPO11-1 (or due to failure of DSB formation), these responses are not induced, so the chromatin and chromosome axes probably remain in a DSB susceptible state. Nevertheless, other meiotic cell cycle modules, such as cyclin-dependent kinase, may still stimulate meiotic progression to ultimately relax the barrier of SC assembly, resulting in synapsis between non-homologous chromosomes in pachytene-like mutant meiocytes.
It has been reported that maize meiocytes during the leptotene-zygotene transition (prezygotene) undergoes dramatic changes, including increased total chromosome volume, decondensed heterochromatic knobs, offset-localized nucleolus and clustering of telomeres [41]. Meiotic DSBs are thought to form at this stage or shortly prior to this transition [68]. Here, we have shown that during this transition the numbers of highly copious SPO11-1 foci inside the nucleus are variable, whereas relatively similar numbers of SPO11-1 foci are loaded on AEs. Interestingly, AEs in wild-type cells gradually became smoother and more linear, which does not occur in spo11-1 meiocytes at the same stage. In addition, the shortening of AEs (from 4 to 3 mm) that we detected in wild-type meiocytes at this transition is likely a response to the presence of SPO11 or DSB formation, but it is not caused by synapsis since ZYP1 signals were detected at later stages in WT meiocytes.
To date, programmed DSB formation in eukaryotic meiosis has been shown to require SPO11. Unlike the single SPO11 protein in mammals and fungi, plants encode at least three SPO11 proteins [69]. In Arabidopsis, genetic analyses revealed that two non-redundant SPO11 proteins are required for meiotic recombination [70][71][72], suggesting that DSBs are catalyzed by a SPO11-1/SPO11-2 heterodimer. Maize has three SPO11 related proteins. Although the meiotic functions of maize SPO11-2 and SPO11-3 are not clear, loss of SPO11-1 alone severely affects DSB formation and homologous recombination. Despite a few DBSs being detected in spo11-1 mutants by our TUNEL assay and γH2AX, RAD51 immunolocalization, these DSBs (or other types of DNA damages) was not repaired by homologous recombination, suggesting that neither SPO11-2 nor SPO11-3 possesses redundant functions for SPO11-1.
Current DSB maps generated by Spo11-oligo sequencing have provided a landscape of recombination initiation sites [63,[73][74][75]. However, despite thousands of DSB hotspots having been identified, only a few hundred DSBs are usually detected during each round of meiosis. Considering the probabilistic nature of DSB formation, further investigations of DSB regulation, distribution and timing at the level of individual cells will likely fill this information gap. Here, our observations based on super resolution microscopy have revealed for the first time in any organism the dynamic localization of SPO11-1 relative to AEs, providing evidence for the 'loop-tethered axis' model. By characterizing the phenotypes of maize spo11-1 mutants, we suggest that there is an intimate relationship between DSB formation and the structure of chromosome axes.

Maize material
Maize inbred lines B73, A344, A632 and Mo17, as well as wild type siblings of the Mu insertional mutants and spo11-1 mutant alleles, were grown in the greenhouse of Academia Sinica and previously in the UC Berkeley greenhouse and experimental field. The mtm99-14 / spo11-1-Δ14 and mtm00-03 / spo11-1-Δ03 mutants came from two independent genetic screens for maize meiotic mutants using a Mutator (Mu) population, respectively, during the 1999 and 2000 field seasons [47,48]. The spo11-1-1 line was found in the TUSC reverse genetic resource [52].

Genetic mapping and Mu insertion determination
To map the mtm00-03 mutation, bulk segregation analysis (BSA) of two mtm00-03 F2 populations created by crossing the mtm00-03 line with inbred A344 or B73 were used, as previously described [76]. Sterile plants showed tight linkage to the polymorphic markers IDP8472 and particularly IDP4279 that map to the end of the short arm of chromosome 5 on Bin 5.01 between physical positions 5.3 and 8.5 Mb. For each mutant, Mu insertion border sequences that are associated with segregation of phenotype were determined using the Mu-Illumina border determination approach, as previously described [51].

RT-PCR analysis
RT-PCR analyses were conducted on RNA extracted using Trizol from meiotic anthers, as described previously [77]. The Afd1 252 base pair (bp) cDNA and PhyC 444 bp cDNA fragments, as well as the 3' and 5' portions of Spo11-1 transcripts, were amplified using specific primers (S7 Table). The PhyC1 and PhyC2 fragments presented as two distinct bands from genomic DNA but with an identical size to cDNA. All PCRs were performed using DMSO and the Invitrogen GoTaq kit with 30 cycles using annealing temperature of 52˚C except for the SPO11-1 5' amplification using 60˚C. Quantitative PCR of Spo11-1α and Spo11-1β transcripts was performed using FastStart Universal SYBR Green Master mix with specific primers (S7 Table). Cyanase was used as an internal normalization control. PCRs were performed on an ABI 7500 real-time PCR system in a 64-well plate. The RT-qPCR reaction was performed in triplicate for each RNA sample. Specificity of the amplification products and Ct values were analyzed using ABI prism dissociation curve analysis software.

Antibody production
To generate anti-SPO11-1 antibody, we used a peptide corresponding to amino acids 9-36 of SPO11-1 protein (RAAPLEGDEQQLRRRLEEAALLLRRIKG) as antigen to generate a rabbit polyclonal antibody, followed by affinity purification (LTK BioLaboratories, Taiwan). Antibody specificity was validated by dot-blot analysis (S7 Fig).

Western blot
Anthers were dissected from immature tassels after their meiotic stages and sizes had been determined and were then snap frozen in liquid nitrogen. Protein extraction was performed as previously described [37]. Two hundred micrograms of proteins were loaded in a 20 cm long, 10% SDS-PAGE gel, run for 18 h at 100 V, and then transferred to a polyvinylidene fluoride (PVDF) membrane. The PVDF membrane was blocked overnight with 5% milk in Tris-buffered saline-Tween (TBST) solution (containing 1X TBS and 0.1% Tween 20) at 4˚C. Subsequently, the PVDF membrane was incubated with anti-SPO11-1 antibody (1:1000 dilution in 5% milk/TBST) and then with horseradish peroxidase conjugated goat anti-rabbit IgG secondary antibody (1:5000 dilution in TBST). After a TBST wash, signals were detected using UVP Biospectrum 600 imaging system.

Immunolocalization
Immature tassels were carefully removed from 4-to 6-week-old plants and kept in moist paper towel until dissection or fixation. Anthers were dissected after their meiotic stages and sizes had been determined and their positions on a tassel were recorded as a relative time course. We analyzed meiosis in synchronized cohorts of meiocytes in anthers using characteristics described previously [78]. Immunostaining was performed as described [79] with primary  1:100). The anti-RAD51 antibody was a gift from Wojciech Pawlowski (Cornell University) and anti-γH2AX was a gift from Hong Ma and Yingxiang Wang (Fudan University). Conjugated secondary antibodies were obtained from Molecular Probes. Slides were imaged using either an LSM microscope 780 with ELYRA equipped with SR-SIM (Zeiss) or a Deltavision core microscope (GE).

Image processing
The super-resolution z-stacking images were examined using ZEN2012 software (Zeiss). All image processing procedures were carried out in "ImageJ-FIJI" software. The SPO11-1 signals were segmented as described in S12 Fig. Chromosome axes were skeletonized as described in S14 Fig. To measure distances between SPO11-1 and chromosome axes, a 3D distance map of DSY2-labeled chromosome axes was created using the plugin "3S Image Suite" (S15 Fig), and 3D measurement of objects and distance information was analyzed using the "3D ROI Manager" tool in 3D Image Suite (S16 Fig). The curvature of chromosome axes was analyzed using the Kappa plugin (S19 Fig). All super-resolution images used for measurement have been deposited in the BioStudies database (http://www.ebi.ac.uk/biostudies) under accession number S-BSST322.

TUNEL assay and TEM
The TUNEL assay for DSB detection was performed as described previously [82]. First, meiocytes were embedded in polyacrylamide pads as described in the immunostaining protocol above. The TUNEL reaction was performed in a humid chamber for 2.5 h at 37˚C with the in Situ Cell Death Detection Kit (Roche) using 100 μL of reaction mixture per slide. Slides were mounted and imaged with a Deltavision core microscope. TEM analysis was performed as described previously [82,83]. Given 94% identity between the PhyC1 and PhyC2 coding regions, both PhyC1 and PhyC2 fragments (asterisks) were amplified from genomic DNA (gDNA) of B73 and A344 WT using primers AR105/ AR106 (S7 Table). 0P represents a negative PCR control lacking DNA template. By RT-PCR, the same primers amplified both transcripts (arrowheads) of identical size from WT leaf (L) and anther (A). However, sequencing results confirmed that only the PhyC1 transcript (arrows) was detected in the mtm99-14 (14) and mtm00-03 (03) mutants. (TIF) S3 Fig. Acetocarmine staining of spo11-1-1 meiocytes during meiosis I. (A) spo11-1-1  To validate our SPO11-1 antibody specificity, dot blot analyses were performed using SPO11-1 antiserum (1:1000 dilution) against synthetic peptides of SPO11-1 antigen, SPO11-2 and SPO11-3 in corresponding regions. Their amino-acid sequences are listed below the dot blots. Serial dilutions of equal amounts (1 μg) of peptide were dotted for detection and blots were imaged using the UVP Biospectrum 600 system with exposure times of 5 or 20 min. (C-D) Preimmune serum (C, prebleed) and anti-immune serum (D) were used to test SPO11-1 antibody in WT meiocytes at early zygotene stage by means of immunofluorescence analysis. (E) Western blot analysis using the affinity-purified SPO11-1 antibody for detection of potential SPO11-1 proteins in total protein extracts of WT and spo11-1-Δ03 (Δ03) anthers. Two weak bands (arrows) of~40 kDa were detected in WT, but these were absent from the mutant sample. The segmentation method for the same cell is demonstrated in S12 Fig. (B) The 3D ROI manager tool is opened and the segmented binary image is imported to add all labeled objects.(C) Examination of selected objects in the segmented binary image using 3D ROI manager.(D) Acquisition of geometric measurements of selected objects, including coordinates of the center of objects (CX, CY, CZ), the borders of objects, their volumes (Vol) and compactness (Comp), as well as the distances from the centers to borders of objects (DC). (E) Examination of selected objects (circled in yellow) from the original gray-scale image. (F) Acquisition of the fluorescence intensity of selected objects from the original gray-scale image. Signal intensity values for the center point (AtCenter), total sum (IntDen), minimal (Min) and the mean (Mean) for the objects are obtained. (G) After incorporating the 3D distance map into the 3D ROI manager, selected objects (circled in yellow) could be examined on the 3D distance map. The segmentation and skeletonization method for chromosome axes of the same cell is demonstrated in S14 Fig.(H) Acquisition of distances (μm) between the centers of selected objects to the closest axes (AtCenter), as well as minimal (Min) and mean (Mean) distances from the borders of objects to the closest axes. Axis-associated SPO11-1 objects are defined when X values (i.e. the minimal distance from AEs to borders of an object) equal zero.