Cotton (Gossypium hirsutum), the most important cash crop for natural textile fibres, meanwhile, represents the fifth largest source of vegetable oil for human consumption in the world. Typically, cottonseed oil contains three major fatty acids: 26% palmitic acid, 15% oleic acid and 58% linoleic acid (Liu et al., 2002). The relatively high level of linoleic acid reduces oxidative stability of cottonseed oil, which can cause rancidity, a short shelf life and production of detrimental trans‐fatty acids (Shockey et al., 2017). Oleic acid has better oxidative stability than linoleic acid due to its monounsaturated nature, so it is considered a reliable and healthy fatty acid.

Microsomal ω‐6 fatty acid desaturase (FAD2) can introduce a carbon–carbon double bond at the Δ12 position of oleic acid to form linoleic acid (Figure 1a). Downregulation of FAD2 via the RNA silencing method has been reported to increase oleic acid content in Arabidopsis and cotton (Liu et al., 2002). However, these transgenic lines cannot be used in any practical way due to consumer concerns about GMO and governmental regulatory issues (Shockey et al., 2017). Recently, the availability of versatile CRISPR/Cas genome editing techniques has allowed scientists to precisely edit the expressions of target genes without T‐DNA insertions (Wang et al., 2020; Zhang et al., 2020). Knockout of FAD2 genes by CRISPR/Cas9 editing resulted in accumulation of about 80% oleic acid in soybean seed (Do et al., 2019), which represent promising example of biotechnological production of high‐oleic acid in other oilseed crops. In this study, we generated high‐oleic, nontransgenic cotton using CRISPR/Cas9 editing techniques for the first time.

The G. hirsutum genome (v. HAU) encodes eight homologs (GhFAD2) of the Arabidopsis FAD2 gene (Figure 1a). Among them, the GhFAD2‐1 homologs had the closest relationship with Arabidopsis FAD2 based on protein sequence similarity (72.30%). Transcription analysis across 22 different tissues indicated that GhFAD2‐1A/D expressed in the developing ovule with higher levels in the ovule at 35 days post‐anthesis (dpa), while GhFAD2‐2A/D highly expressed in the stamen. The expression levels of GhFAD2‐3A/D in the leaf and torus were relatively higher. GhFAD2‐4A/D expressed in all tissues of cotton with relatively higher levels in leaf, pistil and ovule at 20 dpa (Figure 1b). Taken together, these findings suggest that GhFAD2‐1A/D is the key gene determining the fatty acid composition of cottonseed oil.

GhFAD2‐1A and GhFAD2‐1D share the same gene structure, no introns, and contain two conserved domains, one for DUF3474 and one for fatty acid desaturase (Figure 1c). In addition, the two GhFAD2‐1 homologs share 97.40% similarity in their amino acid sequences. We chose one target site followed by the CCA PAM motif located in the DUF3474 domain and another target site followed by the CCG PAM motif located in the fatty acid desaturase domain (Figure 1c). These sgRNAs were designed for targeting the two GhFAD2‐1 homologs and integrated into the vector pRGEB32‐GhU6.9‐NPT II following the method described in our previous report (Wang et al., 2018).

We obtained 35 independent T0 plants via Agrobacterium tumefaciens‐mediated transformation of G. hirsutum genotype Jin668 (Li et al., 2019) (Figure 1d). Among these plants, 25 independent plants were positive transformants due to the presence of NPTII and Cas9 genes (Figure 1e and g). To further investigate the editing profile of these plants, the gene regions of GhFAD2‐1A/D were amplified from genomic DNA of leaves by PCR using gene‐specific primers (Figure 1e and g). The results indicated that gene editing occurred at both target sites and that deletions (69.57%) were more abundant than insertions (Figure 1f). For each target site, single nucleotide insertion/deletion was predominant. A total of 86.84% of the loci were deletions of a ‘C’ in single nucleotide deletion events, while 89.29% of insertions were ‘T’ (46.43%) or ‘A’ (42.86%). As predicted, large fragment deletions of 308 and 309 bp were observed between sgRNA1 and sgRNA2 target sites (Figure 1f). Finally, 19 (76%) of the 25 T0 plants were determined to be mutants generated by the CRISPR/Cas9 system. Notably, 73.68% and 68.42% mutant T0 plants contained homozygous mutation at the sgRNA1 and sgRNA2 target sites, respectively.

Three T0 plants (m1, m20 and m27) were selected to assess the inheritance of the mutations because abundant seeds were harvested from these plants. The T1 seedlings were evaluated by PCR (Figure 1g) and Sanger sequencing (Figure 1h). The results showed all mutations induced by CRISPR/Cas9 were stably inherited to T1 generation. Interestingly, new mutations were observed in T1 plants. At the sgRNA1 target site, the m1‐1 plant showed new mutations with deletion of 41 bp, and the m1‐2 plant gained one new ‘C’ deletion at the sgRNA2 target site (Figure 1h). These complex editing patterns indicated that the Cas9 was active in T1 generation or that the T0 plants contained chimeric mutations, which is consistent with the gene‐editing profile of BnITPK genes in Brassica napus (Sashidhar et al., 2020). We further identified that four edited lines (m1‐2, m1‐3, m20‐2 and m27‐3) were nontransgenic due to the absence of both NPTII and Cas9 genes (Figure 1g). Meanwhile, 19 potential off‐target sites were predicted using the CRISPR‐P (v. 2.0) program (Liu et al., 2017). The possible off‐target mutations in m1‐2 plants were further analysed via PCR and Sanger sequencing and no mutations were found in the potential off‐target sites, indicating the high accuracy of the CRISPR/Cas9 system.

The T1 seeds of the four nontransgenic Cas9 edited lines were subjected to fatty acid analyses. As expected, the cotton lines with knockout of GhFAD2‐1A/D exhibited significant increases in oleic acid at the expense of a large reduction of linoleic acid (Figure 1i). In m20‐2 seeds, oleic acid content was 77.72%, which was 5.58 (p < 0.01) times higher than the average level of 13.94% in wild type (WT), and the level of linoleic acid decreased concomitantly from 58.62% to 6.85% (Figure 1i). Additionally, palmitic acid contents were also significantly reduced in the four mutant lines (Figure 1i). Furthermore, to assess the impact of detected mutations on fibre quality, the fibre length, strength and Micronaire were determined and no changes were observed in the m1‐2 or m1‐3 edited lines (Figure 1j). Finally, the phenotype (Figure 1k), and levels of total oil and stearic acid in the nontransgenic seeds remained unchanged in comparison with the WT (Figure 1l). We also found that the mutant lines were not different from the WT in seed germination under normal condition, which is consistent with the germination results of high‐oleic soybean (Bachleda et al., 2017). These results showed that the knockout of GhFAD2‐1A/D had dramatically improved the quality of cottonseed oil and the high‐oleic trait has no side‐effect on major agronomic traits.

In summary, we successfully generated high‐oleic, nontransgenic mutants in allotetraploid upland cotton. This is the first report of generating high‐oleic source material in allotetraploid cotton via CRISPR/Cas9 editing system. These high‐oleic, nontransgenic mutants provide useful parents in breeding programs to introgress a high‐oleic trait into commercial varieties with other agronomically valuable traits.

同时，棉花（陆地棉）是天然纺织纤维最重要的经济作物，是世界上人类食用植物油的第五大来源。通常，棉籽油包含三种主要脂肪酸：26％的棕榈酸，15％的油酸和58％的亚油酸（Liu等，2002）。相对较高的亚油酸含量会降低棉籽油的氧化稳定性，从而导致酸败，保质期短和产生有害的反式脂肪酸（Shockey等人，2017年）。油酸由于其单不饱和性质而具有比亚油酸更好的氧化稳定性，因此被认为是可靠且健康的脂肪酸。

微粒体ω-6脂肪酸去饱和酶（FAD2）可以在油酸的Δ12位置引入碳-碳双键，形成亚油酸（图1a）。据报道，通过RNA沉默方法下调FAD2可增加拟南芥和棉花中油酸的含量（Liu等，2002）。然而，由于消费者对转基因生物和政府监管问题的担忧，这些转基因品系无法以任何实际方式使用（Shockey等人，2017）。最近，通用的CRISPR / Cas基因组编辑技术的可用性使科学家能够精确地编辑目标基因的表达，而无需插入T-DNA（Wang等，2020 ; Zhang等，2020）。通过CRISPR / Cas9编辑敲除FAD2基因导致大豆种子中积累了约80％的油酸（Do等，2019），这是在其他油料作物中生物技术生产高油酸的有前途的例子。在这项研究中，我们首次使用CRISPR / Cas9编辑技术生成了高油酸，非转基因棉花。

该陆地棉的基因组（诉HAU）编码8个同源物（GhFAD2所述的）拟南芥FAD2基因（图1a）。其中，基于蛋白质序列相似性，GhFAD2-1同源物与拟南芥FAD2的亲缘关系最密切（72.30％）。在22个不同组织中进行的转录分析表明，GhFAD2-1A / D在发育中的胚珠中表达，在花后35天（dpa）时胚珠中的水平更高，而GhFAD2-2A / D在雄蕊中高表达。叶片和环面中GhFAD2-3A / D的表达水平相对较高。GhFAD2-4A / D在20 dpa的叶，雌蕊和胚珠中棉花的所有组织中均有表达（图1b）。综上所述，这些发现表明，GhFAD2-1A / D是决定棉籽油脂肪酸组成的关键基因。

GhFAD2-1A和GhFAD2-1D共享相同的基因结构，没有内含子，并且包含两个保守域，一个用于DUF3474，一个用于脂肪酸去饱和酶（图1c）。此外，两个GhFAD2-1同源物在其氨基酸序列中具有97.40％的相似性。我们选择了一个目标位点，然后是位于DUF3474域中的CCA PAM基序，然后选择了另一个目标位点，然后是位于脂肪酸去饱和酶域中的CCG PAM基序（图1c）。这些sgRNA设计用于靶向两个GhFAD2-1同源物，并按照我们以前的报告中所述的方法整合到载体pRGEB32-GhU6.9-NPT II中（Wang等人，2018年）。

我们通过根癌农杆菌介导的陆地棉基因型Jin668转化获得了35株独立的T0植物（Li等，2019）（图1d）。在这些植物中，由于存在NPTII和Cas9基因（图1e和g），因此有25株独立的植物为阳性转化体。为了进一步研究这些植物的编辑谱，我们使用了GhFAD2-1A / D的基因区域使用基因特异性引物通过PCR从叶片的基因组DNA扩增得到（图1e和g）。结果表明，基因编辑发生在两个目标位点，并且缺失（69.57％）比插入更丰富（图1f）。对于每个靶位点，单核苷酸插入/缺失是主要的。在单核苷酸缺失事件中，共有86.84％的基因座是'C'的缺失，而89.29％的插入是'T'（46.43％）或'A'（42.86％）。如预期的那样，在sgRNA1和sgRNA2目标位点之间观察到308和309 bp的大片段缺失（图1f）。最后，确定了25个T0植物中的19个（76％）是由CRISPR / Cas9系统产生的突变体。值得注意的是，73.68％和68.42％的突变T0植物分别在sgRNA1和sgRNA2目标位点包含纯合突变。

选择了三种T0植物（m1，m20和m27）来评估突变的遗传，因为从这些植物中收获了丰富的种子。通过PCR（图1g）和Sanger测序（图1h）评估了T1幼苗。结果表明，由CRISPR / Cas9诱导的所有突变均稳定地遗传到T1代。有趣的是，在T1植物中观察到新的突变。在sgRNA1目标位点，m1-1植物显示出新的突变，缺失了41 bp，而m1-2植物在sgRNA2目标位点获得了一个新的'C'缺失（图1h）。这些复杂的编辑模式表明Cas9在T1世代中活跃，或T0植物含有嵌合突变，这与甘蓝型油菜（Sashidhar）中BnITPK基因的基因编辑谱一致等人，2020）。我们进一步发现，由于缺少NPTII和Cas9基因，四个编辑品系（m1-2，m1-3，m20​​-2和m27-3）是非转基因的（图1g）。同时，使用CRISPR-P（v.2.0）程序预测了19个潜在的脱靶位点（Liu et al。，2017）。通过PCR和Sanger测序进一步分析了m1-2植物中可能的脱靶突变，在潜在的脱靶位点未发现突变，表明CRISPR / Cas9系统具有很高的准确性。

对四个非转基因Cas9编辑品系的T1种子进行了脂肪酸分析。正如预期的那样，敲除GhFAD2-1A / D的棉线显示出油酸的显着增加，但亚油酸的大量减少却导致了损失（图1i）。在m20-2种子中，油酸含量为77.72％，为5.58（p 相对于野生型（WT）的平均水平13.94％高出<0.01）倍，亚油酸的水平从58.62％下降到6.85％（图1i）。此外，在四个突变体系中，棕榈酸含量也显着降低（图1i）。此外，为了评估检测到的突变对纤维质量的影响，确定了纤维长度，强度和马克隆值，并且在m1-2或m1-3编辑行中未观察到变化（图1j）。最后，与野生型相比，非转基因种子的表型（图1k）以及总油和硬脂酸的含量保持不变（图1l）。我们还发现，在正常条件下，突变株系与野生型在种子萌发方面没有差异，这与高油大豆（Bachleda）的发芽结果一致。等人，2017）。这些结果表明，GhFAD2-1A / D的敲除显着提高了棉籽油的质量，高油性对主要农艺性状没有副作用。

总之，我们成功地在异源四倍体陆地棉上产生了高油酸，非转基因突变体。这是通过CRISPR / Cas9编辑系统在异源四倍体棉花中产生高油酸原料的第一份报告。这些高油酸非转基因突变体为育种计划提供了有用的亲本，以使高油酸性状渗入具有其他农业上有价值的性状的商业品种。