Ryegrass is one of the most important forage crops worldwide. It is the basis for 80% of milk production and 70% of meat production and has major economic importance. Breeding programmes for ryegrass started in the 1920s, and breeders have mainly relied on repeated phenotypic and recently genotypic selection of elite individuals. Although this approach has led to significant improvements in several characters including rust resistance, spring growth and aftermath heading, it tends to be laborious, expensive and time‐consuming, mainly due to gametophyte self‐incompatibility in most ryegrass species (Sampoux et al. , 2011). In order to overcome some of the limitations of traditional introgression and selective breeding, modern methods of mutation induction offer attractive opportunities to target specific genes of interest and directly introduce allelic variability. In the last decade, the Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR‐associated endonuclease 9 (CRISPR/Cas9) system has been extensively used in most crops and is paving the way to precision trait improvements in factors including yield, quality, biotic‐ and abiotic stress resistance and breeding rate (Chen et al. , 2019; Ran et al. , 2017; Wang et al. , 2014). However, this powerful tool for genome editing has not yet been used in ryegrass. Meiosis arose early during the evolution of eukaryotes and is vital for sexual reproduction, not only in relation to genomic stability but also to genetic diversity. Meiotic studies of plants in the areas of crop fertility and genetic variation have important potential agronomical applications. DMC1 (DISRUPTED MEIOTIC cDNA1), initially identified in yeast (Bishop et al. , 1992) as a homolog of the bacterial strand exchange protein RecA, is a crucial meiotic recombinase. Here, we describe the use of the CRISPR/Cas9 system to introduce mutations in LpDMC1 in two species: Italian ryegrass (Lolium perenne ssp. multiflorum ) and perennial ryegrass (Lolium perenne ). We succeeded in obtaining both T0 homozygous and heterozygous mutants, and the T0 null mutants of Italian ryegrass exhibited complete male sterility and severely disordered meiosis with univalents and multivalents appearing at diakinesis.

To see whether mutations could be introduced into ryegrass using the CRISPR/Cas9 system, we generated a sgRNA (TS‐LpDMC1) targeting exon 5 of LpDMC1, with an Xce I restriction enzyme site near the protospacer‐adjacent motif (PAM) for ease of analysis (Figure 1a). Because plant tissue culture and genetic transformation are time‐consuming, we tested the activities of sgRNA in a protoplast transient expression system as described by Shan et al. (2013). TS‐LpDMC1 promoted by TaU6 was co‐introduced with SpCas9 into ryegrass protoplasts by PEG‐mediated transformation. After 40‐ to 48‐h incubation, analysis of genomic DNA using a PCR restriction enzyme digestion assay (PCR/RE) demonstrated the occurrence of indel mutations at the target site (Data not shown).

To determine whether the CRISPR/Cas9 method was applicable to other ryegrass genes, we targeted the ryegrass orthologue of centromere‐specific histone H3 variant (CENH3). In Arabidopsis thaliana , CENH3 ensures faithful transmission of the genome at cell division, and when cenh3 null mutants producing altered CENH3 proteins are crossed with wild type, many haploid Arabidopsis plants are generated (Ravi and Chan, 2010). When we co‐transformed a sgRNA targeting exon 3 of LpCENH3 along with SpCas9 into protoplasts (Figure 1d), PCR/RE analysis revealed frameshift mutations at the target site (Figure 1e). These results show that CRISPR/Cas9 can be used to generate mutations in ryegrass.

Next, the sgRNA expression cassette was combined with SpCas9 in a single DNA construct by GIBSON Assembly and introduced along with a hygromycin‐resistant plasmid into preconditioned embryogenic callus (EC) lines of ryegrass by gold particles bombardment. To generate these EC lines, seeds of Italian ryegrass cultivar Gepetto and perennial ryegrass cultivar Goyave were de‐husked and sterilized, and somatic EC lines were established as described (Ran et al. , 2014). Three separate lines designated Gepetto‐8, Gepetto‐66 and Goyave LMG LDF‐Lp3711 (provided by DLF Seeds) with outstanding regeneration ability were selected for transformation. After bombardment, the EC was transferred to hygromycin medium. Surviving calli were obtained after 4 weeks’ induction and sub‐culture. Thereafter, they were regenerated for 8 weeks; and plantlets with established roots were transferred to potting mix for mutants’ identification. The mean transformation efficiencies of these three lines were 3.83%, 4.50% and 2.66%, respectively. The entire experimental cycle took approximately 10 months from target design to mutant identification. Compared with conventional methods, CRISPR/Cas9 provides a rapid and straightforward method for genetic manipulation of ryegrass.

T0 generation LpDMC1 knockout mutants were identified by PCR/RE using the same primers as in the protoplast assays (Figure 1b), followed by Sanger sequencing. In total, we obtained eight mutants in two species: seven of Italian ryegrass and one of perennial ryegrass. G8‐12 was a homozygous mutant with a 9 bp in‐frame deletion in one allele. The other non‐wild‐type plant G8‐16 was a mosaic, with three mutations: −1 bp, −33 bp deletions and −10/+37 bp deletion/insertion. The other six were heterozygotes (Figure 1c). The genome editing efficiencies of LpDMC1 in Gepetto‐8, Gepetto‐66 and Goyave LMG LDF‐Lp3711 were 11.63%, 11.11% and 5.88%, respectively.

To detect off‐target events, putative off‐target sites of TS‐LpDMC1 were predicted using the draft genome sequence of the forage grass Lolium perenne (Byrne et al. , 2015) (Figure 1f). Two non‐mismatch off‐target sites, LpOff3 and LpOff4, with highly similar surrounding sequences, which could produce mutations in two different non‐coding regions, were selected for further study. PCR products that included LpOff3 /LpOff4 from the eight mutants were sequenced, but no mutations were detected (Figure 1g). The next‐closest match was LpOff5 , which had an imperfect PAM, suggesting that off‐target events would be unlikely (Figure 1f).

Because Italian and perennial ryegrass are both gametophyte self‐incompatibility and have different vernalization requirements, it was difficult and time‐consuming to obtain homozygous T1 mutants. Since disruption of DMC1 in Arabidopsis , rice and barley usually leads to reduced fertility, we examined the pollen grains produced by wild type and the T0 Lpdmc1 mutant G8‐16. Upon iodine‐potassium iodide (I₂‐KI) staining, the pollen grains from the wild type appeared mostly round, while those from G8‐16 were empty and shrunken (Figure 1h,i). This phenotype points to complete male sterility, as in the Osdmc1a Osdmc1b Tos‐17 double insertion mutant of rice (Wang et al. , 2016). The wild type pollen grains did not stain as strongly as those of rice, probably due to the difference in starch content between Italian ryegrass and rice. To clarify what happened during Lpdmc1 meiosis, we stained meiotic chromosomes with 4’,6‐diamidino‐2‐phenylindole (DAPI). The meiotic chromosomes of the Lpdmc1 mutant behaved normally at leptotene (Figure 1j,m). However at diplotene, whereas circular or figure‐of‐eight shapes of the chromatids were observed in wild type (Figure 1k), in the Lpdmc1 mutant the chromatids were stuck together, mostly due to the existence of multivalents (Figure 1n). At diakinesis, unlike the wild type, which formed 7 bivalents (Figure 1l), condensed univalents and multivalents were scattered throughout the mutant nuclei (Figure 1o). These results point to a meiotic defect in the Lpdmc1 mutant.

In summary, we were able to introduce mutations in two genes, LpDMC1 and LpCENH3, of ryegrass in vivo using the CRISPR/Cas9 system. We obtained eight T0 knockout mutants of LpDMC1 in Italian and perennial ryegrass. The Lpdmc1 null mutants were completely male sterile with severely disrupted meiosis, indicating that DMC1, a highly conserved protein, plays a pivotal role in meiosis in many species. We are now using genome editing tools to improve other agronomic traits in ryegrass. We anticipate that this method will come to be used routinely in this economically important forage crop.

黑麦草是世界上最重要的牧草作物之一。它是牛奶产量80％和肉类产量70％的基础，具有重要的经济意义。黑麦草的育种计划始于1920年代，育种者主要依靠对精英个体的重复表型和最近的基因型选择。尽管这种方法已大大改善了一些特性，包括抗锈蚀性，春季生长和后遗症，但由于大多数黑麦草种类中配子体自交不亲和性，它往往费力，昂贵且耗时（Sampoux等，2011年）。为了克服传统基因渗入和选择性育种的某些局限性，现代的突变诱导方法为靶向感兴趣的特定基因并直接引入等位基因变异性提供了诱人的机会。在过去的十年中，成簇的规则间隔的短回文重复序列/ CRISPR相关的核酸内切酶9（CRISPR / Cas9）系统已在大多数农作物中得到广泛使用，并为改善产量，品质，生物和遗传等因素的精确性状铺平了道路。非生物胁迫的抗性和繁殖率（Chen等，2019 ; Ran等，2017 ; Wang等，2014）。但是，这种用于基因组编辑的强大工具尚未在黑麦草中使用。减数分裂发生在真核生物进化的早期，并且对于有性繁殖至关重要，这不仅与基因组稳定性有关，而且与基因多样性有关。作物育性和遗传变异领域的植物减数分裂研究具有重要的潜在农学应用。DMC1（DISRUPTED MEIOTIC cDNA1）是一种至关重要的减数分裂重组酶，最初在酵母菌中被发现（Bishop等，1992），它是细菌链交换蛋白RecA的同源物。在这里，我们描述了使用CRISPR / Cas9系统在LpDMC1中引入两种物种的突变：意大利黑麦草（Lolium perenne ssp。何首乌）和多年生黑麦草（黑麦草）。我们成功获得了纯合和杂合的T0突变体，意大利黑麦草的T0无效突变体表现出完全的雄性不育和严重减数分裂，在诊断时出现单价和多价。

为了查看是否可以使用CRISPR / Cas9系统将突变引入黑麦草中，我们生成了一个靶向LpDMC1外显子5的sgRNA（TS ‐ LpDMC1），在原间隔物相邻基序（PAM）附近有一个Xce I限制酶位点。分析（图1a）。由于植物组织培养和遗传转化非常耗时，因此我们如Shan等人所述在原生质体瞬时表达系统中测试了sgRNA的活性。（2013）。TS ‐ LpDMC1通过PEG介导的转化，将TaU6促成的蛋白与SpCas9共同引入黑麦草原生质体。孵育40到48小时后，使用PCR限制酶消化分析（PCR / RE）对基因组DNA进行分析，发现在目标位点发生插入缺失突变（数据未显示）。

为了确定CRISPR / Cas9方法是否适用于其他黑麦草基因，我们针对着丝粒特异性组蛋白H3变体（CENH3）的黑麦草直系同源基因。在拟南芥中，CENH3确保细胞分裂时基因组的可靠传递，当产生改变的CENH3蛋白的cenh3空突变体与野生型杂交时，会产生许多单倍体拟南芥植物（Ravi和Chan，2010年）。当我们将靶向LpCENH3外显子3的sgRNA与SpCas9一起转化为原生质体时（图1d），PCR / RE分析揭示了目标位点的移码突变（图1e）。这些结果表明，CRISPR / Cas9可用于在黑麦草中产生突变。

接下来，通过GIBSON Assembly将sgRNA表达盒与SpCas9组合成单个DNA构建体，并与抗潮霉素的质粒一起通过金粒子轰击将其引入黑麦草的预处理胚性愈伤组织（EC）品系中。为了产生这些EC系，对意大利黑麦草品种Gepetto和多年生黑麦草品种Goyave的种子进行去壳和灭菌处理，并按照所述方法建立体细胞EC系（Ran等，2014）。选择三个具有优异再生能力的分离系Gepetto-8，Gepetto-66和Goyave LMG LDF-Lp3711（由DLF Seeds提供）进行转化。轰击后，将EC转移至潮霉素培养基中。经过4周的诱导和继代培养，获得了存活的愈伤组织。此后，将它们再生8周。将根已确定的小植株转移到盆栽混合物中以鉴定突变体。这三条线的平均转化效率分别为3.83％，4.50％和2.66％。从靶标设计到突变体鉴定，整个实验周期耗时约10个月。与传统方法相比，CRISPR / Cas9为黑麦草的遗传操作提供了一种快速而直接的方法。

使用与原生质体检测相同的引物通过PCR / RE鉴定T0代LpDMC1敲除突变体（图1b），然后进行Sanger测序。总共，我们获得了两个物种的八个突变体：七个意大利黑麦草和一个多年生黑麦草。G8-12是一个纯合突变体，在一个等位基因中有9 bp的读框缺失。另一种非野生型植物G8-16是一个镶嵌体，具有三个突变：−1 bp，−33 bp缺失和−10 / + 37 bp缺失/插入。其他六个是杂合子（图1c）。Gepetto-8，Gepetto-66和Goyave LMG LDF- Lp3711中LpDMC1的基因组编辑效率分别为11.63％，11.11％和5.88％。

为了检测脱靶事件，使用草类多年生黑麦草的基因组序列草案预测了TS ‐ LpDMC1的脱靶位点（Byrne等，2015）（图1f）。选择了两个非错配脱靶位点LpOff3和LpOff4，它们具有高度相似的周围序列，这些位点可能在两个不同的非编码区产生突变，需要进一步研究。对包括来自八个突变体的LpOff3 / LpOff4的PCR产物进行了测序，但未检测到突变（图1g）。下一场最接近的比赛是LpOff5，其PAM不够完善，表明脱靶事件不太可能发生（图1f）。

由于意大利黑麦草和多年生黑麦草都是配子体自交不亲和的，且具有不同的春化要求，因此获得纯合的T1突变体既困难又费时。由于在拟南芥，水稻和大麦中破坏DMC1通常会导致生育力降低，因此我们研究了野生型和T0 Lpdmc1突变体G8-16产生的花粉粒。碘-碘化钾（I ₂ -KI）染色后，来自野生型的花粉粒大部分呈圆形，而来自G8-16的花粉粒则空而缩小（图1h，i）。这种表型表明男性完全不育，就像水稻的Osdmc1a Osdmc1b Tos-17双插入突变体一样（Wang等，2016）。野生型花粉粒不像大米那样强染色，这可能是由于意大利黑麦草和大米之间的淀粉含量不同所致。为了阐明在Lpdmc1减数分裂过程中发生了什么，我们用4'，6-二mid基-2-苯基吲哚（DAPI）对减数分裂染色体进行了染色。Lpdmc1突变体的减数分裂染色体在瘦素下表现正常（图1j，m）。然而，在双线，而圆形或数字的8字形在野生型（图1K）观察染色单体的形状，在Lpdmc1突变的染色单体粘在一起，主要是由于存在多价（图1n）。在诊断中，与野生型不同，野生型形成7个二价键（图1l），缩合的单价和多价键分散在整个突变核中（图1o）。这些结果表明Lpdmc1突变体的减数分裂缺陷。

总之，我们能够使用CRISPR / Cas9系统在体内将黑麦草的两个基因LpDMC1和LpCENH3引入突变。我们在意大利和多年生黑麦草中获得了八个LpDMC1的T0敲除突变体。该Lpdmc1空突变体是完全不育与严重破坏减数分裂，表明DMC1，一个高度保守的蛋白质，起着在许多物种在减数分裂了举足轻重的作用。我们现在正在使用基因组编辑工具来改善黑麦草的其他农艺性状。我们预计这种方法将在这种具有重要经济意义的饲料作物中常规使用。