Resistant starch (RS) refers to starch that is not digested in the stomach or small intestine, and provides health benefits by reducing glycaemic index and promoting gut health (Hazard et al., 2020). Barley (Hordeum vulgare L.) is the fourth most widely cultivated cereal worldwide, and there is growing interest in barley as a healthy food. The RS content of cereal grains is positively associated with the presence of amylose and long-chain amylopectin (Li et al., 2021). However, increasing the levels of amylose in crops is still challenging. Overexpressing the enzyme involved in amylose synthesis, granule-bound starch synthase (GBSS), does not increase amylose content in most species (Seung, 2020). By contrast, suppressing amylopectin synthesis enzymes such as starch synthase (SS) and starch-branching enzyme (SBE) isoforms in cereals increases amylose content significantly (Chen et al., 2021). However, knockout mutations in these genes usually compromise the yield potential. Recent progress in CRISPR/Cas9-mediated gene editing make it possible to induce targeted mutations in multiplex genes (Cheng et al., 2023; Lawrenson et al., 2015; Luo et al., 2021), which provides a promising approach for devising new strategies to increase amylose content while avoiding/minimizing production limitations. We reasoned that multiplex editing could allow the contribution of all other SS and SBE isoforms to amylose content to be systematically assessed, and allow screening for optimum combinations of mutations that could synergistically provide strong increases in amylose content.

Single-guide RNAs (sgRNAs) were designed to target the exons of seven desired genes encoding four SSs (SSI, SSIIa, SSIIIa and SSIV) and three SBEs (SBEI, SBEIIa and SBEIIb) based on the genome sequence of barley cv. Golden Promise (Table S1). The CRISPR/Cas9 vector was introduced into immature embryos of Golden Promise using Agrobacterium-mediated transformation (Method S1). We identified 113 edited plants from 152 T₀ transformants, and these plants contained mutations in one to six of the targeted genes; however, no transgenic plants contained mutations in all seven target genes as no SSI mutant plants were identified (Figure 1a; Table S2). Sequencing showed that the nucleotide mutations in the target regions led to premature stop codons, which would disrupt the functional structures of the corresponding proteins (Figure 1b).

In the T₃ generation, we obtained 10 edited lines with different genotypes that were homozygous for mutations in one to three target genes. Western blotting of grain protein extracts from these lines showed that the target proteins were absent or dramatically reduced in abundance, consistent with the mutations disrupting normal protein accumulation (Figure 1c; Figure S1–S10). Notably in the ssIIa mutant, in addition to the absence of SSIIa, the abundance of SSI, SBEIIa and SBEIIb was also undetectable or strongly reduced relative to the wild type (WT). This is consistent with observations in ssIIa mutants of other cereal crops and can be explained by the formation of a multienzyme complex of SSI/SSIIa/SBEIIa or SBEIIb (Liu et al., 2012). Grains from the mutant lines exhibited varying degrees of shrunken morphology, and the SSIIa mutation induced the greatest negative effect on grain weight among the five monogenic mutants (Figure 1a). Interestingly, combining the SSIIa and SSIIIa mutations appeared to overcome this grain weight deficiency (Table S3).

Scanning electron microscopy revealed that the starch granules from the edited lines exhibited a dramatically altered morphology with hollow, concave surface and sticky or amorphous shapes (Figure 1d). Significantly, the ssIIassIIIassIVa and ssIIasbeIIasbeIIb lines had a few sizable A-type granules that contained several small, concave granules with or without clear boundaries. The particle size distribution of starch granules showed that starches from the sbeIIa, sbeIIb and ssIIassIIIa mutants contained significantly more A-type granules but fewer B-type granules than the controls, while the other mutants had fewer A-type granules and more B-type granules than the controls (Figure 1e).

The total starch contents of the 10 lines were significantly lower than that of the controls, and they all except ssIIassIVa and ssIIassIIIassIVa had significantly higher starch contents than the ssIIa mutant (Figure 1f; Table S4). Relative to the controls, apart from sbeIIb, which expressed a slight but not significant increase in the amylose content, the amylose contents of all mutants were greatly increased. Significantly, the ssIIasbeIIasbeIIb and sbeIIasbeIIb lines exhibited the highest amylose content (87.43% and 86.82%, respectively), which were around fourfold higher than control lines and 1.5-fold higher than the ssIIa mutant. Fluorescence-activated capillary electrophoresis indicated that inactivating the starch synthase genes also dramatically affected the chain-length distribution of amylopectin (Figure 1e). Furthermore, the thermal properties, swelling power and solubility of the starches in all mutants were striking changed compared with the controls (Figure 1g).

Most of the edited lines contained greatly elevated contents of RS, β-glucan, fructan and fibre than the controls (Figure 1f; Table S4). In particular, the polygenic mutants sbeIIasbeIIb and ssIIasbeIIasbeIIb had extremely high RS (12.27% and 14.50%, respectively), which were 35-fold greater than that of the controls. The ssIIassIIIassIVa, ssIIasbeIIasbeIIb and ssIIassIVa mutants exhibited the highest fructan, β-glucan and fibre contents among all mutants, respectively; they were greater than that of the controls and the ssIIa.

In summary, we successfully used multiplex editing to generate barley mutants harbouring single, double and triple mutations in five starch synthesis genes and produced either higher dietary fibre content or higher grain weights than those previously achieved by employing a single gene target (i.e., SSIIa). Three mutants, ssIIassIVa, sbeIIasbeIIb and ssIIasbeIIasbeIIb, exhibited improved levels of amylose and dietary fibre and/or a higher grain weight compared with the single ssIIa mutant. These were determined to be the best choices among the five polygenic mutants for future applications in the production of healthy food products. Our study demonstrates that Cas9-mediated multiplex gene editing is feasible for modifying starch to generate grain with higher RS contents than targeted editing of single genes and provides a theoretical basis and genetic resource for breeding barley with improved health benefits.

抗性淀粉（RS）是指不在胃或小肠中消化的淀粉，通过降低血糖指数和促进肠道健康来提供健康益处（Hazard等，  2020）。大麦 ( Hordeum vulgare L.) 是全球第四大种植最广泛的谷物，人们对大麦作为健康食品的兴趣与日俱增。谷物中的 RS 含量与直链淀粉和长链支链淀粉的存在呈正相关（Li et al .,  2021）。然而，提高农作物中的直链淀粉水平仍然具有挑战性。过度表达参与直链淀粉合成的酶，即颗粒结合淀粉合酶 (GBSS)，不会增加大多数物种的直链淀粉含量 (Seung,  2020 )。相比之下，抑制谷物中的支链淀粉合成酶，例如淀粉合成酶（SS）和淀粉分支酶（SBE）同工型，会显着增加直链淀粉含量（Chen等人，  2021）。然而，这些基因的敲除突变通常会损害产量潜力。CRISPR/Cas9介导的基因编辑的最新进展使得在多重基因中诱导靶向突变成为可能（Cheng等人，  2023；Lawrenson等人，  2015；Luo等人，  2021），这为设计提供了一种有前途的方法。增加直链淀粉含量同时避免/最小化生产限制的新策略。我们推断，多重编辑可以系统地评估所有其他 SS 和 SBE 同工型对直链淀粉含量的贡献，并允许筛选可以协同地大幅增加直链淀粉含量的突变的最佳组合。

单引导 RNA (sgRNA) 被设计为靶向编码四个 SS（SSI、SSIIa、SSIIIa 和 SSIV）和三个 SBE（SBEI、SBEIIa 和 SBEIIb）的七个所需基因的外显子，这些基因基于大麦品种的基因组序列。黄金承诺（表 S1）。使用农杆菌介导的转化（方法S1）将CRISPR/Cas9载体引入Golden Promise的未成熟胚胎中。我们从 152 个 T ₀转化体中鉴定出 113 株经过编辑的植物，这些植物含有 1 到 6 个目标基因的突变；然而，没有转基因植物在所有七个目标基因中都含有突变，因为没有发现 SSI 突变植物（图 1a；表 S2）。测序表明，目标区域的核苷酸突变导致终止密码子过早出现，这会破坏相应蛋白质的功能结构（图1b）。

在 T ₃代中，我们获得了 10 个具有不同基因型的编辑品系，这些品系对于一到三个目标基因的突变是纯合的。对这些品系的谷物蛋白提取物进行蛋白质印迹分析表明，目标蛋白不存在或丰度显着降低，这与破坏正常蛋白积累的突变一致（图 1c；图 S1-S10）。值得注意的是，在ssIIa突变体中，除了不存在 SSIIa 之外，SSI、SBEIIa 和 SBEIIb 的丰度也相对于野生型 (WT) 检测不到或大大减少。这与其他谷类作物的ssIIa突变体的观察结果一致，并且可以通过 SSI/SSIIa/SBEIIa 或 SBEIIb 多酶复合物的形成来解释（Liu等，2012）。突变系的籽粒表现出不同程度的萎缩形态，在五个单基因突变体中， SSIIa突变对籽粒重量的负面影响最大（图 1a）。有趣的是，结合SSIIa和SSIIIa突变似乎可以克服这种粒重缺陷（表 S3）。

扫描电子显微镜显示，编辑后的品系中的淀粉颗粒表现出显着改变的形态，具有中空、凹面和粘性或无定形形状（图1d）。值得注意的是，ssIIassIIIassIVa和ssIIasbeIIasbeIIb系具有一些相当大的 A 型颗粒，其中包含一些带有或不带有清晰边界的小凹面颗粒。淀粉颗粒粒径分布表明，与对照相比，sbeIIa、sbeIIb和ssIIassIIIa突变体淀粉中A型颗粒明显增多，B型颗粒明显减少，而其他突变体A型颗粒较少，B型颗粒较多颗粒比对照（图 1e）。

10个品系的总淀粉含量显着低于对照，并且除ssIIassIVa和ssIIassIIIassIVa外，所有品系的淀粉含量均显着高于ssIIa突变体（图1f；表S4）。相对于对照，除了sbeIIb直链淀粉含量略有增加但不显着外，所有突变体的直链淀粉含量均大幅增加。值得注意的是，ssIIasbeIIasbeIIb和ssIIasbeIIb品系表现出最高的直链淀粉含量（分别为87.43％和86.82％），比对照品系高约四倍，比ssIIa突变体高1.5倍。荧光激活毛细管电泳表明，淀粉合酶基因失活也显着影响支链淀粉的链长分布（图1e）。此外，与对照相比，所有突变体中淀粉的热性质、膨胀力和溶解度都发生了显着变化（图1g）。

大多数编辑品系的 RS、β-葡聚糖、果聚糖和纤维含量比对照显着升高（图 1f；表 S4）。特别是，多基因突变体sbeIIasbeIIb和ssIIasbeIIasbeIIb具有极高的RS（分别为12.27％和14.50％），比对照高35倍。ssIIassIIIassIVa 、ssIIasbeIIasbeIIb 和 ssIIassIVa突变体分别在所有突变体中表现出最高的果聚糖、β-葡聚糖和纤维含量；它们比对照和ssIIa更大。

总之，我们成功地使用多重编辑产生了五个淀粉合成基因中含有单、双和三突变的大麦突变体，并产生了比以前通过使用单基因靶标（即 SSIIa）实现的更高的膳食纤维含量或更高的谷物重量。 。与单个ssIIa突变体相比，三个突变体ssIIassIVa、sbeIIasbeIIb和ssIIasbeIIasbeIIb表现出改善的直链淀粉和膳食纤维水平和/或更高的粒重。这些被确定为未来应用于健康食品生产的五种多基因突变体中的最佳选择。我们的研究表明，Cas9介导的多重基因编辑对于修饰淀粉以产生比单基因靶向编辑更高RS含量的谷物是可行的，并为育种提高健康效益的大麦提供了理论基础和遗传资源。