Meiotic chromosome axis remodelling is critical for meiotic recombination in Brassica rapa

Abstract Meiosis generates genetic variation through homologous recombination (HR) that is harnessed during breeding. HR occurs in the context of meiotic chromosome axes and the synaptonemal complex. To study the role of axis remodelling in crossover (CO) formation in a crop species, we characterized mutants of the axis-associated protein ASY1 and the axis-remodelling protein PCH2 in Brassica rapa. asy1 plants form meiotic chromosome axes that fail to synapse. CO formation is almost abolished, and residual chiasmata are proportionally enriched in terminal chromosome regions, particularly in the nucleolar organizing region (NOR)-carrying chromosome arm. pch2 plants show impaired ASY1 loading and remodelling, consequently achieving only partial synapsis, which leads to reduced CO formation and loss of the obligatory CO. PCH2-independent chiasmata are proportionally enriched towards distal chromosome regions. Similarly, in Arabidopsis pch2, COs are increased towards telomeric regions at the expense of (peri-) centromeric COs compared with the wild type. Taken together, in B. rapa, axis formation and remodelling are critical for meiotic fidelity including synapsis and CO formation, and in asy1 and pch2 CO distributions are altered. While asy1 plants are sterile, pch2 plants are semi-sterile and thus PCH2 could be an interesting target for breeding programmes.


Introduction
Meiosis is a specialized cell division in sexually reproducing organisms shuffling maternal and paternal genomes through homologous recombination (HR) and independent assortment of homologous chromosomes. HR is initiated by programmed formation of SPO11catalysed DNA double-strand breaks (DSBs) (Keeney and Kleckner, 1995;Keeney et al., 1997;Neale et al., 2005;Pan et al., 2011). DSBs undergo a series of transitions, being repaired differently (Osman et al., 2011;Mercier et al., 2015;Wang and Copenhaver, 2018) into either a crossover (CO; reciprocal genetic exchange between homologous chromosomes) or a non-crossover (NCO), using as repair template either the sister chromatid or the homologous chromosome exchanging only short stretches of DNA. Erroneous meiotic DSB formation or repair may lead to univalent formation or chromosome breakage and thus genome instability, reducing fertility. In plants <10% of DSBs are repaired as COs, and DSBs/COs occur heterogeneously along chromosomes (Muyt et al., 2009;Mercier et al., 2015) limiting genetic variation generated during each meiotic cycle. In many crop species, COs are restricted to chromosome ends, inhibiting access to traits residing in 'cold' regions or creating 'linkage drag'; therefore, modulating CO number and distribution is of interest for plant breeding (Phillips et al., 2013;Choulet et al., 2014;Li et al., 2015;Demirci et al., 2017;Lambing and Heckmann, 2018;Dreissig et al., 2019).
Concurrently with HR, the meiotic nucleus undergoes extensive reorganization of the chromatin. Following S phase, sister chromatids are linked by cohesins and, during leptotene, are organized in chromatin loops tethered to a linear proteinaceous structure called the meiotic chromosome axis (Kleckner, 2006). In Arabidopsis, this structure comprises, for example, ASY1, ASY3, ASY4, SMC3, and REC8 (Caryl et al., 2000;Cai et al., 2003;Chelysheva et al., 2005;Lam et al., 2005;Ferdous et al., 2012;Chambon et al., 2018). DSBs form at preferred sites (hotspots) in the chromatin loops which are then tethered to the axes where they are repaired as COs or NCOs (Panizza et al., 2011). Following DSB formation, a proteinaceous structure, called the synaptonemal complex (SC), starts to form during zygotene by progressively polymerizing between homologous chromosome axes physically connecting them and promoting CO formation (Page and Hawley, 2004).
To study the role of axis remodelling in a crop species, we characterized asy1 and pch2 mutants in Brassica rapa (2n=20). asy1 plants form axes but SC formation is defective, leading to a reduction in CO frequency. Interestingly, residual chiasmata are proportionally increased towards chromosome ends and are enriched in the major 45S rDNA-carrying chromosome; in particular, chiasmata form within or close to the 45S rDNA locus while this is not the case in the wild type (WT) or in pch2. pch2 plants form a partial SC while axes are not remodelled; that is, ASY1 loading during leptotene is reduced and ASY1 is not depleted from synapsed regions, leading to a reduction in CO frequency. PCH2-independent chiasmata are more skewed towards terminal chromosome regions. To corroborate this cytological observation, in Arabidopsis CO rates were established in different genetic intervals, revealing increased CO rates in telomeric chromosome regions at the expense of decreased (peri-)centromeric CO rates in pch2 compared with the WT. We conclude that ASY1 and PCH2 are critical for SC and CO formation during meiosis in B. rapa and that asy1 and pch2 show altered CO patterning. Due to pch2 showing only semi-sterility, it could potentially be an interesting target in breeding programmes to redistribute COs.

Genotyping
In B. rapa, PCR-based genotyping was performed using derived cleaved amplified polymorphic sequences (dCAPS) (Neff et al., 1998) with primers selected using the dCAPS Finder 2.0 tool (http://helix.wustl.edu/ dcaps/) (Neff et al., 2002). A list of primers and restriction enzymes is shown in Supplementary Table S1. After PCR amplification, the resulting amplicons were digested with corresponding restriction enzymes at 37 °C overnight and resolved on 2.5% agarose gels. Primers used to genotype Arabidopsis T-DNA mutants are indicated in Supplementary  Table S1.

RNA extraction and reverse transcription-PCR
Total RNA of sample tissues was extracted with an RNeasy Plant Mini Kit (Qiagen) performing DNase digestion with the RNase-Free DNase Set (Qiagen). Reverse transcription was performed in a 20 µl reaction employing 750 ng of total RNA with a Tetro cDNA synthesis kit (Bioline) using oligo(dT) primers. Expression of Actin was evaluated as technical control for the integrity of the RNA/cDNA. A 5 µl aliquot of undiluted cDNA was used as template for 36 (PCH2-12) and 28 (Actin) PCR cycles. Primers used are listed in Supplementary Table S1.

Cytology and microscopy
Cytological procedures were carried out as described (Armstrong et al., 2009) with minor modifications. 5S (pCT4.2; Campell et al., 1992) and 45S (pTa71; Gerlach and Bedbrook, 1979) rDNA fluorescence in situ hybridization (FISH) probes were labelled by nick translation with Texas Red and Atto488 (NT labelling kits, Jena Biosciences). Chromosome spreading for immunostaining in B. rapa was done using five anthers per slide; digestion was done during 8 min in a moist chamber at 37 °C, disrupting the material with a brass rod after the first 4 min; and spreading was done with 1.5% lipsol. Immunostaining in Arabidopsis was performed on acid chromosome spreads from male meiocytes as described (Lambing et al., 2020). The following antibodies and dilutions were used: anti-ASY1 (rabbit/rat, Sanchez-Moran et al., 2007, 1:500), anti-ZYP1 (rat/guinea pig, Higgins et al., 2005, 1:500), anti-SMC3 (rat, Ferdous et al., 2012, 1:300), anti-ASY3 (rabbit, Ferdous et al., 2012, 1:200), anti-MLH1 (for B. rapa rabbit, Jackson et al., 2006; for Arabidopsis rabbit, Chelysheva et al., 2010, 1:200), and anti-HEI10 (rat, Lambing et al., 2015, 1:200). ASY1 intensity in B. rapa asynapsed versus synapsed regions was measured as described . Quantification of ASY1 signal intensity in Arabidopsis leptotene nuclei was performed according to Ziolkowski et al. (2017). In brief, Arabidopsis images were acquired as z-stacks of 10 images with an optical section of 0.2 µm each. The maximum fluorescence intensity projection for each cell was generated, and total signal intensity was quantified using ImageJ. A region adjacent to the chromatin was used to quantify the background level that was deducted from the total signal intensity. Each signal intensity was normalized by the mean WT signal intensity for comparison. Quantification of ASY1 signal intensity in B. rapa leptotene nuclei was performed in a similar way, but images were acquired as a single plane, and measurements were done with the corresponding microscope software. Images were acquired with a Nikon Eclipse 90i fluorescence microscope equipped with a Nikon DS-Qi1Mc digital camera and NIS-Elements-F software, or a Nikon Eclipse Ni-E equipped with a Nikon DS-Qi2 camera and NIS-Elements-AR version 4.60 software (Nikon, Tokyo, Japan). Images were processed with GIMP 2.10 (www.gimp.org).
For PCH2, one line with a predicted premature stop codon and a second line with an SNP in an intron-exon boundary were selected (Fig. 1A): pch2-9 (662G>A, W143>Stop), predicted to encode 29% of the native protein length, and pch2-12 (694G>A, intron4-exon4 boundary). SNPs were confirmed by sequencing. Reverse transcription-PCR revealed the presence of alternative splice variants in pch2-12 ( Supplementary  Fig. S1). Although the presence of the PCH2 WT transcript cannot be excluded, pch2-12 probably results in a null mutation as pch2-12 is allelic to pch2-9 in terms of meiotic fidelity (see below). PCH2 expression is found in flower buds (containing cells undergoing meiosis) and also in leaves, similar to the expression pattern found for other meiotic genes (Caryl et al., 2000;Grelon et al., 2001).
For all lines, three backcrosses to the WT (R-o-18) were performed in order to decrease the secondary mutation load. Data presented correspond to plants from the third backcross generation.  Table S2). To explore whether reduced fertility was due to meiotic defects, we performed male meiotic chromosome spread analysis.
During WT meiosis ( Fig. 1D; Supplementary Fig. S2a), unpaired chromosomes appear as thin threads during leptotene. During zygotene, chromosomes start to align and synapse, reaching full synapsis visible as thick threads at pachytene. At diakinesis, homologous chromosomes are visible as bivalents physically connected by chiasmata (cytological manifestation of COs). At metaphase I, 10 bivalents align at the equatorial plate and 10 homologous chromosomes migrate to opposite poles during anaphase I. During the second meiotic division, chromosomes align at metaphase II and 10 chromatids separate to opposite poles during anaphase II, producing four haploid products in a balanced tetrad.
In asy1, thick chromatin threads indicative of synapsis were not observed (Fig. 1D). During diakinesis and metaphase I, most chromosomes appeared as univalents, indicating a failure to form COs. During anaphase I, univalents either segregated to one pole or showed precocious separation of sister chromatids (Fig. WT, asy1-13 (a13), asy1-14 (a14), pch2-9 (p9), and pch2-12 (p12). (C) Average, SD, and range of bivalents, univalents, and chiasmata per cell in analysed lines. n=number of cells; a minimum of two independent plants per genotype. (D) B. rapa WT, asy1, and pch2 male meiotic chromosome spreads. In both mutants, synapsis and chiasma formation are impaired, leading to the occurrence of univalents at diakinesis that results in unequal chromosome segregation and unbalanced tetrads. Note, univalents in the mutants could migrate complete to a pole (arrowheads) or separate chromatids (asterisks) during anaphase I. DNA is counterstained with DAPI and shown in grey. Scale bar=10µm.
In asy1, except for #3, all discernible chromosomes including the largest and the smallest appeared as univalents at similar frequencies (79-86%) (Supplementary Table S3). However, #3 which possesses the NOR with the largest 45S block, appeared as univalent in only 37% of cells. From the 63% #3 bivalents, 84% were rods and 16% rings; and from the #3 rod bivalents, in 64% of asy1 cells the chiasma was cytologically associated with the 45S rDNA; that is, the chiasma was found either inside the NOR or distal to it (Fig. 2B). In contrast to asy1, in WT and pch2 #3 rod bivalents, the single chiasma was never associated with the NOR but instead was invariably formed in the opposite arm (Fig. 2B).

Chromosome axes form in asy1 but fail to assemble a synaptonemal complex
The asy1 mutants encode <25% of the native protein length, thus having a truncated or deleted HORMA domain and closure motif, critical for ASY1 localization (West et al., 2019;Yang et al., 2020b). To determine if the truncated ASY1 protein localizes to the axes and an SC forms in the asy1 cells, we performed ASY1 and ZYP1 (transverse filament protein of the SC; Higgins et al., 2005) immunolocalization. In the WT (Fig. 3), ASY1 localized to chromosome axes during leptotene. Once ASY1 was fully polymerized, ZYP1 initially formed foci and then short stretches which progressively elongated until all homologous chromosomes were fully synapsed. In asy1 (Fig.  3), ASY1 could not be detected. ZYP1 predominantly formed foci (from <5 to >20, possibly in a stage-dependent manner) and rarely elongated beyond forming short stretches, indicating that SC formation is largely impaired. In late pachytene-/early diplotene-like cells, ZYP1 formed aggregates probably due to its inability to assemble the SC (Zickler and Kleckner, 1999).
To analyse axis morphogenesis in asy1, we performed immunolocalization of the cohesin subunit SMC3 (Lam et al., 2005) and the coiled-coil protein ASY3 (Ferdous et al., 2012). In the WT (Fig. 4), SMC3 was present from leptotene to pachytene, co-localizing initially with ASY1 and later on with ZYP1. Upon synapsis, ASY1 signals became faint while SMC3 persisted, gradually thickening as synapsis proceeded. ASY3 initially localized to the nucleolus during leptotene similar to Arabidopsis (Ferdous et al., 2012) before it co-localized with ASY1 and then with ZYP1; however, ASY3 signals were patchier and less linear than those of ZYP1. In asy1 (Fig. 4), both SMC3 and ASY3 localized to the axis following WT dynamics. Fig. 3. Synapsis is defective in asy1 and pch2. Immunolocalization of ASY1 (red) and ZYP1 (green) in the WT, asy1, and pch2. In asy1, ASY1 is not detected and ZYP1 forms foci in zygotene-pachytene cells and aggregates in late pachytene-diplotene cells. In pch2, different from the WT, ASY1 does not get depleted from the axes and appears highly abundant in limited synapsed regions co-localizing with ZYP1. DNA is counterstained with DAPI and shown in blue. Scale bar=10µm. Thus, ASY1-independent axes form but are insufficient to support SC formation, resulting in reduced CO formation.

Class I and class II crossovers are reduced in asy1
CO formation was reduced but not abolished in asy1. From 178 asy1 cells analysed, only one cell showed 20 univalents, whereas in all others between one and eight chiasmata were observed. To evaluate the nature of the remaining COs in asy1, we analysed the chiasma frequency distribution per cell. In the WT, the majority of COs are sensitive to interference, which leads to a non-random numerical distribution between cells. As a result, the mean chiasma frequency significantly deviates from a Poisson distribution [χ 2 (R-o-18) =46.32, P<0.0001, n=94] (Fig. 5). However, in asy1, the chiasma frequency per cell was not significantly different from a Poisson distribution [χ 2 (asy1-14) =7.79, P=0.25, n=82] (Fig. 5), suggesting that residual ASY1independent COs are randomly distributed between cells.
To better understand CO formation in asy1, we performed immunolocalization of the class I CO marker MLH1 (Jackson et al., 2006) together with ZYP1. We scored on average 9.44±0.72 (n=37) and 8.7±4.79 (n=10) MLH1 foci per cell in the WT and asy1, respectively (Fig. 6A). The MLH1 foci number per cell was highly variable in asy1 (1-15) but not in the WT (8-11). While in the WT MLH1 foci typically overlapped with ZYP1, in asy1 only an average of 3.9 MLH1 foci co-localized with residual ZYP1 foci (range 0-9). This manifested in different ways: ZYP1 stretches appeared to originate from MLH1 foci, two MLH1 foci were 'bridging' a ZYP1 stretch, or an MLH1 focus formed between two ZYP1 foci (Fig. 6A). Assuming that only ZYP1-associated MLH1 foci or a subset of these are CO competent, at least some of the residual chiasmata in asy1 could be ZMM-dependent class I COs.
We also immunolocalized HEI10 Wang et al., 2012) together with ZYP1. In the WT, numerous small HEI10 foci found during early pachytene progressively fade, leaving on average 11.31±1.71 bright HEI10 foci per cell (n=27, range 8-15) during diplotene-diakinesis (Fig. 6B). In asy1, this number was significantly lower and more variable (9±3.3, n=19, range 4-14). This variability could be attributed to the difficulty in cytologically defining meiotic prophase stages in asy1 and it is possible that some of the observed HEI10 foci will not mature in CO sites. However, as in the WT, in asy1, HEI10 foci typically co-localized with ZYP1. Similar to MLH1, short ZYP1 stretches originated from HEI10 foci or two HEI10 foci bridged a ZYP1 stretch. Altogether, our data suggest that interference-sensitive COs are designated in asy1, but the synaptic defect compromises CO maturation. As a result, an obligate CO fails to form between all chromosome pairs, and those COs that do mature exhibit a random distribution.
Next, we asked whether class II COs also form in the absence of ASY1 by taking advantage of available Arabidopsis resources. We crossed Arabidopsis asy1 either with shoc1 (involved in class I CO formation; Macaisne et al., 2008Macaisne et al., , 2011 or with mus81 (involved in class II CO formation; Berchowitz et al., 2007;Higgins et al., 2008a), and compared the chiasma frequency in these lines. Under our growth conditions, Arabidopsis asy1 showed 2.3±0.88 bivalents and 2.77±1.22 chiasmata per cell (Table 1), shoc1 showed 2.38±1.04 bivalents and 2.79±1.41 Fig. 4. ASY3 and SMC3 localize to chromosome axes in asy1. Immunolocalization of SMC3 (red, left) and ASY3 (red, right) together with ASY1 (green, WT top) and ZYP1 (green, WT bottom and asy1). ASY3 and SMC3 display similar dynamics in WT and asy1 cells. DNA is counterstained with DAPI and shown in blue. Scale bar=10µm. chiasmata per cell, and mus81 invariably showed five bivalents. In asy1/shoc1 and in asy1/mus81, fewer chiasmata formed than in asy1; in asy1/shoc1, we found a reduction of ~72% and in asy1/mus81 a reduction of ~21% compared with asy1. This suggests that residual COs in asy1 are both class I and II.

Axis remodelling is defective in pch2
In pch2, ASY1 loaded onto chromosomes and, once ASY1 was fully polymerized, ZYP1 appeared initially as foci and later on as stretches. However, full ZYP1 polymerization was never observed (Fig. 3). Indeed, measuring ZYP1 extension in WT and pch2-9 nuclei revealed a large variation in the extension of the SC across pch2 cells and, on average, a reduction of 62% in SC length (Mann-Whitney-Wilcoxon test, P<0.0001, n=13 WT and n=31 pch2) (Fig. 7).
Furthermore, in contrast to the WT, in pch2 ASY1 remained brightly stained following ZYP1 installation (Fig. 3), suggesting that ASY1 is not depleted from chromosome axes at synapsed regions. In various organisms including Arabidopsis, PCH2 is needed for the programmed removal of ASY1 during zygotene from chromosome axes concomitant with SC installation (Börner et al., 2008;Martinez-Perez et al., 2008;Wojtasz et al., 2009;Lambing et al., 2015). To validate this observation, we measured ASY1 intensity (Supplementary Table S4) and found that in the WT there was a 55% reduction in ASY1 intensity in synapsed versus asynapsed regions, whereas in pch2 there was a 21% increase. This increase in pch2 is probably based on the juxtaposition of the two homologous chromosome axes, leading to an increase in signal intensity per pixel.
Interestingly, in asynapsed regions, ASY1 intensity was consistently lower in pch2 than in the WT. In Arabidopsis, PCH2 is involved not only in ASY1 removal, but also in its loading (Yang et al., 2020b). To check whether ASY1 loading is also defective in pch2, we measured ASY1 intensity in leptotene whole nuclei. Despite substantial variation in ASY1 intensity values among slides/experiments, ASY1 intensity was consistently lower in pch2 than in the WT (range 14-76%). To corroborate this cytological observation, we also immunostained ASY1 in Arabidopsis WT and pch2, and found that ASY1 signal intensity was significantly reduced at leptotene in Arabidopsis pch2 (Mann-Whitney-Wilcoxon test, P=4.01 e-03 , n=24 Col-0 and n=24 pch2) ( Supplementary Fig. S3). In summary, our data indicate that PCH2 in B. rapa is important for initial ASY1 loading during leptotene and programmed removal of ASY1 from chromosome axes concomitant with SC installation during zygotene.

Crossover patterning in pch2: distalized crossover at the expense of interstitial crossover
In pch2, chiasma frequency was highly variable and in ~50% of cells the obligatory CO was lost. To better understand how CO control is altered, we analysed CO frequency distribution. We found that chiasma frequency in pch2 did not significantly deviate from a Poisson-predicted distribution [χ 2 (pch2-9) =14.70, P=0.26, n=98)] (Fig. 5), revealing a random distribution of chiasmata between cells. To identify whether interference-sensitive class I COs form in pch2, we performed immunolocalization of HEI10 together with ZYP1 (Fig. 6B). In pch2-9, we found an average of 10.5±3.15 (range 4-15) HEI10 foci during diplotene-diakinesis. These values are comparable with the average chiasma number in pch2-9 (9.97±3.29, range 3-17). Interestingly, we found nuclei where several HEI10 foci localized in close proximity onto one ZYP1 stretch (Fig. 6B, asterisks) together with ZYP1 stretches devoid of HEI10 foci. Thus, a majority of COs in pch2 showing a random distribution according a Poisson-predicted distribution probably belong to class I COs (marked by HEI10 foci).
We next asked whether residual chiasmata in pch2 are positioned differently compared with the WT. We scored the number of terminal versus interstitial chiasmata and found that the percentage of interstitial chiasmata among all chiasmata was 8% in pch2 and 14% in the WT (Fig. 8A, B). Similarly, in asy1, among residual chiasmata, only 6% were interstitial while the majority were found towards chromosome ends (Fig. 8B). Thus, cytologically residual chiasmata in pch2 and asy1 are proportionally more frequently located in terminal chromosome regions compared with the WT. To corroborate our cytological finding on altered chiasma patterning in pch2 with proportionally increased terminal and decreased interstitial chiasmata, we asked whether in Arabidopsis pch2 a similar situation occurs. We re-analysed published recombination data from six genetic intervals  and extended this study by analysing Fig. 6. Class I CO formation in asy1 and pch2. Immunolocalization of MLH1 (A, red) and HEI10 (B, red) together with ZYP1 (green) in the WT, asy1, and pch2. In asy1, a subset of MLH1/HEI10 foci co-localize with ZYP1 foci/short stretches (arrows). In pch2, several HEI10 signals appear close to each other in a single ZYP1 stretch (asterisks). Note, to depict HEI10 dynamics in the WT, cells are shown ranging from zygotene (left) to diakinesis (right), whereas in all other cases representative examples of cells used for quantification are shown. DNA is counterstained with DAPI and shown in blue. Scale bars=5µm. recombination frequency in three more genetic intervals. We classed the genetic intervals based on their chromosomal locations as subtelomeric, interstitial, and centromeric. We found that subtelomeric intervals show increased recombination frequency, while interstitial intervals show few changes, and centromeric intervals show a drastic reduction in recombination frequency in pch2 (Fig. 8C). These data suggest that COs are repressed in the heterochromatic centromeric regions and that they are distalized in the absence of PCH2. Immunostaining of MLH1 revealed that total MLH1 foci are reduced in pch2 (10.4 versus 9.0, Mann-Whitney-Wilcoxon test, P<0.01, n=41 WT and n=20 pch2) and that the reduction is mostly seen on the DAPI-dense regions that are representative of the heterochromatic (peri-)centromeric regions (1.7 versus 0.8, Mann-Whitney-Wilcoxon test, P<0.001, n=41 WT and n=20 pch2) (Fig. 8D). Together, these data reveal that PCH2 is required for CO patterning control.

Discussion
In this work, we have characterized asy1 and pch2 in B. rapa. Both mutants display reduced chiasma formation and univalents at metaphase I that lead to unbalanced gametes and reduced fertility, highlighting the importance of the axis and SC morphogenesis for CO formation. Depletion of ASY1 or PCH2 strongly impacts SC formation, resulting in differential patterning of residual chiasmata being proportionally enriched towards chromosome ends.  Average, SD, and range of bivalents, chiasmata, and ring bivalents per cell in Arabidopsis asy1, shoc1, mus81 single mutants and asy1/shoc1 and asy1/ mus81 double mutants. Note: the bivalent frequency between asy1 and asy1/mus81 was not significantly different, while the number of ring bivalents in asy1/mus81 was reduced, causing the observed reduction in chiasma frequency. n=number of cells.

ASY1-independent meiotic chromosome axes are insufficient for synaptonemal complex and crossover formation
Our analyses indicate that asy1-13 and asy1-14 are allelic and probably represent null mutations. Similar to asy1 in other plant species (Armstrong et al., 2002;Nonomura et al., 2004), B. rapa asy1 is largely, if not completely, asynaptic (Figs 1D,  3). ZYP1 forms foci and occasionally short stretches which are commonly distorted in appearance (Fig. 3), probably representing ZYP1 aggregates (Zickler and Kleckner, 1999) or polycomplexes forming at sites where ZYP1 polymerization/ extension along lateral elements is impaired due to a defect in ASY1-dependent interhomologue bias and recombination progression (Higgins et al., 2005). Despite the absence of an SC, residual chiasmata form in B. rapa asy1 (Fig. 1C, D). This is reminiscent of Arabidopsis and the corresponding hop1 mutant in budding yeast (Hollingsworth Fig. 8. Distalization of chiasmata in pch2 and in asy1. (A) Representative example of B. rapa diakinesis with a majority of chiasmata being interstitial (asterisks) in the WT (left) and distal (arrowheads) in pch2 (right). DNA is counterstained with DAPI and shown in grey. Scale bar=5 µm. (B) Quantification of interstitial versus total chiasmata in B. rapa WT, pch2, and asy1 reveals reduced interstitial CO frequency in the mutants. (C) Recombination frequency in Arabidopsis at genetic intervals defined with genes coding for fluorescent proteins. Genetic intervals are classed based on their locations on the chromosomes. Asterisks represent statistically significant differences (generalized linear model) and 'n.s.' means non-significant. (D) Left: staining of MLH1 (red) and DAPI (blue) at diakinesis in Arabidopsis WT and pch2. Scale bars=10 µM. Right: plots showing total MLH1 foci count, MLH1 foci count on euchromatin, and foci count on heterochromatin per cell. Error bars represent the SD. Asterisks represent statistically significant differences (Mann-Whitney-Wilcoxon test) and 'n.s.' means non-significant. and Johnson, 1993; Ross et al., 1997) but contrasts with rice pair2 (Nonomura et al., 2004). Axis mutants asy3 or asy4 in Arabidopsis also show reduced COs (but to a lesser extent than asy1), and asy1 is epistatic to them in terms of chiasma frequency (Ferdous et al., 2012;Chambon et al., 2018), suggesting that ASY1 is needed upstream from ASY3/ASY4 to regulate DMC1 dynamics and thus interhomologue recombination (Sanchez-Moran et al., 2007). Moreover, ASY4 is required for axial organization of ASY1 and ASY3 in Arabidopsis (Chambon et al., 2018), and ASY3 for correct ASY1 localization in Arabidopsis, maize, and rice (Wang et al., 2011;Ferdous et al., 2012;Lee et al., 2015), while cytologically ASY3 and the cohesin SMC3 localize correctly in asy1 (Sanchez-Moran et al., 2007;Ferdous et al., 2012). Similarly, in B. rapa, SMC3 and ASY3 display WT-like axes localization in asy1 (Fig. 4), suggesting WT-like dynamics of the core axisassociated proteins despite subsequent synaptic and CO defects.
In Arabidopsis axis mutants, the majority of the remaining COs are dependent on the class I CO (ZMM) pathway similar to the WT; for example, in asy1, MLH1 foci form on bivalents (Lambing et al., 2020), in asy3/msh4 double mutants chiasma formation is abolished (Ferdous et al., 2012), or in asy4/zip4 and asy4/msh5 double mutants bivalent formation is reduced by ~95% compared with asy4 (Chambon et al., 2018). In B. rapa asy1, both HEI10 and MLH1 foci numbers scored around pachytene-/diplotene-like stages were highly variable compared with the WT, exceeding the number of cytologically scored chiasmata. This observation could be attributed to different causes. First, HEI10 foci numbers vary significantly through prophase I and therefore variation could stem from the difficulty in selecting late prophase cells due to complete absence of synapsis. Second, interestingly, in Arabidopsis asy1, dmc1 or haploid meiosis MLH1 foci were reported on univalent chromosomes, possibly representing sites of intersister repair (Cifuentes et al., 2013;Lambing et al., 2020). Similarly, HEI10 has been indicated to mark intersister events (Voelkel-Meiman et al., 2012) and it follows initial WT dynamics in mutants with severely impaired CO formation such as the ZMM mutant mer3 (Wang et al., 2012). Third, non-co-localizing HEI10 and MLH1 foci were found in B. napus allohaploids (Grandont et al. 2014). Finally, MLH1 in Sordaria is important for interlock resolution (Storlazzi et al., 2010), and although the presence of MLH1 at interlocks has not been shown, this could arise in asy1 due to defective SC polymerization. Assuming that only ZYP1-associated ZMM foci or a subset of these are CO competent (Ferdous et al., 2012), then probably only a proportion of observed foci will mature into class I CO sites.
In B. rapa asy1, chiasmata were strongly reduced but not abolished, and cytologically foci corresponding to class I CO markers HEI10 and MLH1 formed, suggesting that residual chiasmata may arise via the class I CO pathway. In support of this, by using available Arabidopsis resources, we showed that a majority of ASY1-independent chiasmata are dependent on the ZMM pathway, while a subset is also MUS81 dependent, indicative of class II CO formation (Table 1). Interestingly, as in the WT, roughly similar proportions of class I and class II COs arise in the Arabidopsis asy1 mutant, suggesting that both CO pathways are similarly dependent on ASY1. In future, to determine if the same applies in B. rapa, it will be interesting to evaluate CO formation in asy1/msh4 (class I CO-defective mutant with ~4 residual chiasmata; Blary et al., 2018).
Residual chiasmata in asy1 are proportionally enriched in terminal chromosome regions and on the major 45S rDNA-carrying chromosome Residual chiasmata in B. rapa asy1 are proportionally enriched towards chromosome ends and, except for the major 45S rDNAcarrying chromosome, all discernible chromosomes appear as univalents with similar frequencies, suggesting no positive correlation between chromosome size and chiasma formation.
Terminal dominance of residual chiasmata was also found in Arabidopsis asy1 and further axis-associated mutants such as asy3 and asy4 (Sanchez-Moran et al., 2007;Ferdous et al., 2012;Chambon et al., 2018). It has been proposed that this could be a consequence of recombination initiating by telomeric regions (Sanchez-Moran et al., 2007) and that ASY1 could be antagonizing this telomere-led recombination (Lambing et al., 2020). This highlights the key role of axis components in regulating CO frequency and distribution.
Notably, residual chiasmata showed a strong bias towards the 45S rDNA-carrying chromosome and, in particular, towards the 45S rDNA chromosome end (Fig. 2). Due to limited cytological resolution, we cannot distinguish whether chiasmata involved ribosomal repeats or rather formed in nonribosomal repeat-free regions distal to the 45S rDNA or intermingled with it. We prefer that CO enrichment is found in non-ribosomal repeat-free chromosome ends, as bivalents were typically linked via what seems to be cytologically a very terminal chiasma, and also because we could not see any signs of repair/chromosome segregation problems which would be likely to arise if recombination would occur within highly repetitive regions due to the possibility of non-allelic exchanges (Sasaki et al., 2010). This would agree with recent data from Arabidopsis indicating that exclusion of ASY1 from the NOR in chromosomes 2 and 4 early during prophase is implicated in restricting DSB formation and thus HR in the NOR (Sims et al., 2019). Nonetheless, chiasma enrichment towards the 45S rDNA-carrying chromosome (arm) has also been observed in Arabidopsis asy1 (Sanchez Moran et al., 2001;Lambing et al., 2020). Thus, in asy1, the rDNA probably promotes CO formation in the 45S rDNA-carrying chromosome (arm) (Sanchez Moran et al., 2001;Lambing et al., 2020).
pch2 plants fail to remodel ASY1 and only achieve partial synapsis leading to reduced chiasma formation Both pch2-9 and pch2-12 appear allelic in terms of meiotic behaviour and probably represent null mutants. However, despite three backcrosses, compared with the WT, a slight reduction in fertility was found in pch2-12 segregating families independent of PCH2, suggesting secondary mutation load impacting overall plant fertility. Thus, data were acquired on pch2-9.
In B. rapa pch2, full synapsis was never observed (Figs 1D, 3) despite substantial variation in SC extension (based on ZYP1 immunolocalization) and, on average, SC length was ~62% reduced compared with the WT (Fig. 7). This reduction in SC length is similar to pch2 in Arabidopsis (~68%; Lambing et al., 2015) but differs from rice pch2/crc1, where SC formation is completely abolished (Miao et al., 2013). Notably, in rice pch2, no chiasmata form, while in B. rapa pch2 chiasmata are reduced by ~40% and in Arabidopsis pch2 by ~30%. Moreover, observed chiasma numbers are variable across cells, and often univalents occurred in cells with at least 10 chiasmata, suggesting a defect in CO assurance. Overall, despite species-specific differences, PCH2 in plants is critical for WT levels of synapsis and CO formation.
Brassica rapa pch2 plants form a partial SC while axes are not remodelled; that is, ASY1 loading during leptotene is reduced and ASY1 is not depleted from synapsed regions. This dual function of PCH2 in B. rapa is consistent with data from Arabidopsis and budding yeast (Börner et al., 2008;Lambing et al., 2015;Yang et al., 2020a, b), showing that PCH2 is critical for chromosomal localization of ASY1 as well as for depletion of ASY1 from synapsed regions. In Arabidopsis, in addition to PCH2, the axis proteins ASY3 and ASY4 are also critical for WT dynamics and correct localization of ASY1; for example, in asy4, ASY1 is not depleted from synapsed regions and ASY3 is involved in correct ASY1 recruitment to the chromosome axes (Ferdous et al., 2012;Chambon et al., 2018). Thus, a complex interplay between axis components and associated proteins exists, and not only axis formation but also axis morphogenesis is critical for HR, including DSB repair, template choice, SC formation and, ultimately, CO formation.

Redistribution of chiasmata in pch2
Residual COs in B. rapa pch2, contrary to Arabidopsis , do not significantly deviate from a Poissonpredicted distribution which could be interpreted as suggesting that they are interference insensitive, typical for class II COs. However, the fact that the chiasma frequency distribution does not differ from a Poisson distribution only means that chiasma formation in the mutant has a high random component. In fact, HEI10 foci marking putative class I CO sites appear in pch2 with similar numbers as cytologically scored chiasmata (minimum CO number), with some nuclei having HEI10 foci closely spaced along ZYP1 stretches and at the same time ZYP1 stretches devoid of HEI10 foci (Fig. 6). Possibly maturation of designated COs might be compromised in B. rapa pch2 due to the defect in chromosome axis remodelling, resulting in a CO deficit as in Arabidopsis pch2 . Notably, the average reduction in SC length (62%) is larger than the average reduction in chiasmata (~40%), suggesting that similar to Arabidopsis , the mean reduction in CO frequency was not coordinated with that in SC length. A similar phenotype has been observed in Arabidopsis mutants for the kinesin PSS1 and the E1 enzyme of the neddylation complex AXR1 (Duroc et al., 2014;Jahns et al., 2014), both being strongly defective for synapsis, showing univalents at metaphase I, and with chiasma frequencies similar to the WT but COs being redistributed with closely spaced HEI10 foci along residual synapsed regions.
Furthermore, PCH2-independent chiasmata are cytologically skewed towards terminal chromosome regions in B. rapa. To support this cytological observation, in Arabidopsis pch2, CO rates were established in different genetic intervals, revealing increased CO rates in telomeric chromosome regions at the expense of decreased (peri-)centromeric CO rates compared with the WT (Fig. 8). Thus, considering terminal initiation of SC formation (Hurel et al., 2018), recombination events on limited extended ZYP1 stretches in pch2 might be proportionally enriched in distal regions, explaining the skewed chiasma distribution.
Altogether, in B. rapa, a majority of residual PCH2independent COs are likely to be ZMM-dependent class I COs that are redistributed, being proportionally enriched towards chromosome ends.

Conclusions
Altering CO frequency and distribution is of special interest for plant breeding, since in many crop species CO numbers are limited and skewed towards chromosome ends, limiting recombination and access to naturally available genetic variation and creating linkage drag.
We have analysed mutants of axis-associated ASY1 and axis-remodelling PCH2. Both mutants show reduced CO formation due to defects in meiosis, but interestingly COs are redistributed and at least pch2 plants produce some viable seeds. Considering this altered patterning of chiasmata that arise via the class I and class II CO pathways in asy1 and pch2, it would be interesting to determine whether in combination with HEI10 overexpression (increasing class I CO) or hyperrec mutants (increasing class II COs) . ASY1-/ PCH2-independent chiasmata could be increased, thus potentially increasing bivalent formation and seed setting while maintaining an altered recombination pattern. In any case, considering their general influence on DSB and CO formation including DNA repair template choice, axis-associated or remodelling proteins such as ASY1 or PCH2 are interesting targets to modify meiotic recombination landscapes in the context of plant breeding.

Supplementary data
The following supplementary data are available at JXB online. Table S1. Primers used in this study. Table S2. Plant fertility in the B. rapa WT, and asy1 and pch2 mutants. Table S3. Relative occurrence (%) of 5S/45S-labelled chromosomes as either univalents or bivalents. Table S4. ASY1 immunofluorescence relative signal intensity in WT and pch2 cells. Fig. S1. Alternative splicing of PCH2 in pch2-12. Fig. S2. Male meiotic chromosome behaviour in B. rapa. Fig. S3. Localization defect of ASY1 in Arabidopsis pch2.