Bamboo is a special grass to human due to its great economic and ecological values. Around 2.5 billion people are directly producing and consuming bamboo, and its international trade reached 68.8 billion US dollars in 2018 (Data from International Bamboo and Rattan Organization). One major bamboo species in Asia is Ma bamboo (Dendrocalamus latiflorus Munro ), which is a hexaploid species with three subgenomes (2n  = 72, AABBCC; Guo et al. , 2019). Despite its agronomic importance, it is nearly impossible to modify bamboo traits by traditional breeding as it takes over 70 years to flower. Bamboo research largely lagged behind due to the lack of efficient genetic manipulation tools.

The CRISPR (Clustered Regularly Interspaced Short Palindromic Repeat)/Cas9 provides straightforward ways for genome editing in many plants (Yin et al. , 2017), but has never been applied in bamboo. Here, we reported the generation of bamboo mutants with CRISPR/Cas9 technology by targeting one specific copy or all homoeologous genes.

Since our recently established genetic transformation protocol is time‐consuming (~1.5 years; Ye et al. , 2017), we optimized the CRISPR/Cas9 system in bamboo protoplast. We first improved the protoplast preparation methods and could isolate 3.0 x 10⁶ protoplasts/g fresh leaves. Next, we improved the PEG‐mediated transformation method and reached efficiencies of 53.3% for a single plasmid and 29.8% for two cotransformed plasmids (Figure 1A), which is sufficient for optimizing the CRISPR/Cas9 system. The maize UBI promoter was used to drive Cas9 expression (Ye et al. , 2017). Three polymerase III‐dependent promoters from rice (OsU6a/OsU6b/OsU6c ) were selected to express the sgRNA cassettes (Ma et al. , 2015), as bamboo exhibits high genomic similarity with rice (Peng et al. , 2013). To check the effectiveness of CRISPR/Cas9 constructs, a frameshift mutated GFP (mGFP ) containing an additional ‘guanine’ thereby produces no fluorescence signal was simultaneously cotransformed with CRISPR/Cas9 plasmids (Figure 1B). Around 1.8% of the protoplasts transformed with the UBI‐Cas9/OsU6b‐sgRNA construct showed strong signals within 72 h, indicating that the mGFP function was restored by the CRISPR/Cas9 system through deleting the additional ‘guanine’ (Figure 1C). The OsU6a and OsU6c promoters work as well, however, with lower efficiency than the OsU6b promoter, as positive signals were only occasionally observed with more than 10 replicates. Taking together, the UBI ‐Cas9/OsU6b ‐sgRNA construct effectively works in bamboo protoplast and was used for the following endogenous gene editing in Ma bamboo.

The putative phytoene synthase (PSY1 ) in bamboo, whose homolog in maize functions in carotenoid biosynthesis (Zhu et al. , 2016), was selected for the initial test. Three bamboo PSY1 alleles (DlmPSY1‐A, DlmPSY1‐B and DlmPSY1‐C ) were identified and cloned by a homology cloning strategy (Figure 1D). To mutate all copies of DlmPSY1 , sgRNA1 targeting a conserved site among all DlmPSY1 loci was designed (Figure 1D). In addition, the sgRNA2 target site containing 2–3 single‐nucleotide polymorphisms (SNPs) in the spacer region among three DlmPSY1 homoeoalleles was selected to test the tolerance of sgRNA mismatches (Figure 1D).

A total of 1600 bamboo calluses induced from stem were transformed as described previously (Ye et al. , 2017). In total, 34 independent transgenic lines were confirmed positive (2.1%) by PCR. Based on Sanger sequencing results, 22 (100%) and 10 (83.3%) independent T0 lines were edited in the sgRNA1 and sgRNA2 regions, respectively (Figure 1E), indicating that both constructs effectively induce endogenous gene editing.

The editing profiles were further analysed by sequencing. Eighteen lines (81.8%) contained putative homozygote/biallelic mutations in all subgenomes at the sgRNA1 target site. In some lines, putative homozygote/biallelic mutations exist in one subgenome, while heterozygote or chimeric mutations appear in other subgenomes (T0‐10 and T0‐26; Figure 1F). Eight mutation types were identified from 590 independent clones (Figure 1G). The most frequent mutation type was deletion (75%), of which 59.1% are small deletions (<2 bp). The ratios of large fragment deletions (≥14 bp), insertions and combined indels were 15.9%, 2.21% and 7.82%, respectively (Figure 1G). Since bamboo propagates through asexual budding, those homozygote/biallelic mutations will remain in the genome of their offspring clones during breeding.

sgRNA2 that perfectly targets DlmPSY1‐A1, but not DlmPSY1‐B1 or DlmPSY1‐C1 was designed to study the recognition specificity (Figure 1D). Sequencing results confirmed that 10 transgenic lines contain mutations in DlmPSY1‐A1 , but none in DlmPSY1‐B1 and DlmPSY1‐C1 (Figure 1E). Two lines (20%) were putative homozygous or biallelic mutations (T0‐12 and T0‐14), while 7 lines (70%) were heterozygous/chimeric (T0‐30 to T0‐32 as representative examples, Figure 1H). The ratios of deletions, insertions and combined mutations were 86%, 9% and 5%, respectively (Figure 1I). The mutations were predominantly short nucleotide changes (1–26 bp), and 22.7% were 1bp nucleotide deletions (Figure 1I). Those data demonstrated the successful application of the CRISPR/Cas9 system in mutating a specific DlmPSY1 allele.

Eighteen lines (81.8%) with homozygote/biallelic mutations in all subgenomes at the sgRNA1 site exhibited albino phenotypes (Figure 1J), which appeared at an early stage during tissue culture and persisted at the plantlets stage (Figure 1J). Those results suggest that genome editing takes place at an early stage in embryonic cells and led to the loss of function of all DlmPSY1 alleles. Similar results were reported in rice, wheat or cotton (Wang et al. , 2014; Wang et al. , 2018; Zhang et al. , 2014). In case of sgRNA2, although DlmPSY1 ‐A was mutated, no visible phenotypic change was observed due to the existence of the wild‐type DlmPSY1‐B and DlmPSY1‐C alleles.

Next, we applied this technology in bamboo molecular research. Bamboo is the tallest grass in the world, while the underlying mechanism is unknown. Previously, we identified several Gibberellin‐responsive genes including GRG1 (GA‐responsive gene 1, PH01004823G0070 ) that potentially acts in controlling bamboo height (Zhang et al. , 2018). Here, two homozygote grg1 mutants (efficiency 40%) in Ma bamboo were produced using our optimized CRISPR/Cas9 technology. Mutation in GRG1 increased plant height ( Figure 1K), mostly due to elongated internodes (Figure 1L‐N). Sequencing results confirmed that the grg1 mutant has the putative homozygous mutation in A1 subgenome, biallelic mutation in B1 subgenome and homozygous mutation in C1 subgenome (Figure 1O), indicating the loss of function of GRG1 in transgenic bamboo. To our knowledge, this is the first example on controlling bamboo height through gene manipulation, which will contribute to subsequent studies on the molecular mechanisms behind the fast growth of bamboo.

In summary, for the first time we engineered the hexaploid Ma bamboo through CRISPR/Cas9 technology. The homozygote mutations were obtained in the first generation of transgenic lines, which are extremely important for bamboo species due to its long vegetative growth periods. We also confirmed the albino phenotype of dlmpsy1 mutant in bamboo and generated a bamboo mutant with altered plant height. This demonstrates the applicability of CRISPR/Cas9 in bamboo and thereby boosts future bamboo research and breeding.

竹子具有巨大的经济和生态价值，是人类特有的草。大约有25亿人直接生产和消费竹子，2018年其国际贸易达到688亿美元（国际竹藤组织的数据）。亚洲的一种主要竹种是马竹（Dendrocalamus latiflorus Munro），它是具有三个亚基因组的六倍体物种（2 n  = 72，AABBCC; Guo et al。，2019）。尽管它具有重要的农艺学意义，但要花70多年才能开花，通过传统育种几乎不可能改变竹子的性状。由于缺乏有效的基因操作工具，Bamboo研究在很大程度上落后了。

CRISPR（聚簇的有规律间隔的短回文重复序列）/ Cas9为许多植物的基因组编辑提供了直接的方法（Yin等人，2017），但从未在竹子中应用。在这里，我们报道了通过靶向一种特定拷贝或所有同源基因，利用CRISPR / Cas9技术生成竹突变体。

由于我们最近建立的遗传转化方案非常耗时（〜1.5年; Ye et al。，2017），因此我们优化了竹原生质体中的CRISPR / Cas9系统。我们首先改进了原生质体的制备方法，可以分离出3.0 x 10 ^6个原生质体/ g新鲜叶片。接下来，我们改进了PEG介导的转化方法，单个质粒的效率达到53.3％，两个共转化质粒的效率达到29.8％（图1A），这足以优化CRISPR / Cas9系统。玉米UBI启动子用于驱动Cas9表达（Ye等，2017）。水稻的三种依赖聚合酶III的启动子（OsU6a / OsU6b / OsU6c由于竹子与水稻具有高度的基因组相似性（Peng et al。，2013），因此选择）来表达sgRNA盒（Ma et al。，2015）。为了检查CRISPR / Cas9构建体的有效性，将含有额外``鸟嘌呤''的移码突变GFP（mGFP）从而不产生荧光信号，同时将其与CRISPR / Cas9质粒共转化（图1B）。用UBI-Cas9 / OsU6b-sgRNA构建体转化的原生质体中约有1.8％在72小时内显示强信号，表明CRISPR / Cas9系统通过删除额外的``鸟嘌呤''恢复了mGFP功能（图1C）。这OsU6a和OsU6c启动子也可以工作，但是其效率比OsU6b启动子低，因为仅偶尔观察到阳性信号重复10次以上。综上所述，UBI- Cas9 / OsU6b- sgRNA构建体可在竹原生质体中有效发挥作用，并用于马竹中的以下内源基因编辑。

最初的测试选择了竹中假定的植绿体合酶（PSY1），其在玉米中的同源性在类胡萝卜素生物合成中起作用（Zhu等人，2016）。通过同源克隆策略鉴定并克隆了三个竹PSY1等位基因（DlmPSY1-A，DlmPSY1-B和DlmPSY1-C）（图1D）。为了突变DlmPSY1的所有拷贝，设计了靶向所有DlmPSY1基因座中保守位点的sgRNA1（图1D）。此外，选择了三个DlmPSY1同等位基因间隔区中包含2–3个单核苷酸多态性（SNP）的sgRNA2目标位点，以测试sgRNA错配的耐受性（图1D）。

如前所述（Ye et al。，2017），共转化了1600个由茎诱导的老茧。通过PCR，总共确认了34个独立的转基因品系阳性（2.1％）。根据Sanger测序结果，分别在sgRNA1和sgRNA2区域中编辑了22条（100％）和10条（83.3％）独立的T0系（图1E），表明这两种构建体均有效地诱导了内源基因编辑。

通过测序进一步分析编辑概况。十八个品系（81.8％）在sgRNA1目标位点的所有亚基因组中均包含推定的纯合子/双等位基因突变。在某些品系中，一个亚基因组中存在推定的纯合子/双等位基因突变，而其他亚基因组中则存在杂合子或嵌合突变（T0-10和T0-26；图1F）。从590个独立克隆中鉴定出8种突变类型（图1G）。最常见的突变类型是缺失（75％），其中59.1％是小缺失（<2 bp）。大片段缺失（≥14bp），插入和合并插入缺失的比率分别为15.9％，2.21％和7.82％（图1G）。由于竹子通过无性芽繁殖，因此在育种过程中，那些纯合子/双等位基因突变将保留在其后代克隆的基因组中。

设计完美靶向DlmPSY1-A1而不靶向DlmPSY1-B1或DlmPSY1-C1的sgRNA2来研究识别特异性（图1D）。测序结果证实10个转基因品系在DlmPSY1-A1中包含突变，但在DlmPSY1-B1和DlmPSY1-C1中没有突变。（图1E）。两条品系（20％）是假定的纯合子或双等位基因突变（T0-12和T0-14），而7品系（70％）是杂合/嵌合的（T0-30至T0-32是代表性实例，图1H）。缺失，插入和组合突变的比例分别为86％，9％和5％（图1I）。突变主要是短核苷酸变化（1-26 bp），而22.7％是1bp核苷酸缺失（图1I）。这些数据证明了CRISPR / Cas9系统在突变特定DlmPSY1等位基因方面的成功应用。

在sgRNA1位点所有亚基因组中具有纯合子/双等位基因突变的18个品系（81.8％）表现出白化病表型（图1J），其在组织培养期间出现并在苗期持续（图1J）。这些结果表明，基因组编辑发生在胚胎细胞的早期，并导致所有DlmPSY1等位基因的功能丧失。在水稻，小麦或棉花上也报道了类似的结果（Wang等，2014； Wang等，2018； Zhang等，2014）。对于sgRNA2，尽管DlmPSY1 - A突变，由于存在野生型DlmPSY1-B和DlmPSY1-C等位基因而未观察到表型变化。

接下来，我们将该技术应用于竹子分子研究。竹子是世界上最高的草，但其潜在机制尚不清楚。以前，我们鉴定了几种赤霉素响应基因，包括GRG1（GA响应基因1，PH01004823G0070），它们可能在控制竹子高度方面发挥作用（Zhang等，2018）。在这里，使用我们优化的CRISPR / Cas9技术生产了麻竹中的两个纯合子grg1突变体（效率40％）。GRG1的突变增加了株高（图1K），这主要是由于节间伸长（图1L-N）所致。测序结果证实grg1该突变体在A1亚基因组中具有假定的纯合突变，在B1亚基因组中具有双等位基因突变，在C1亚基因组中具有纯合突变（图1O），表明GRG1在转基因竹中的功能丧失。据我们所知，这是第一个通过基因操纵来控制竹子高度的例子，这将有助于随后对竹子快速生长背后的分子机制的研究。

总而言之，我们首次通过CRISPR / Cas9技术设计了六倍体的Ma Bamboo。纯合子突变是在第一代转基因品系中获得的，由于其营养生长期长，因此对于竹种而言极为重要。我们还证实了dlmpsy1突变体在竹子中的白化表型，并产生了具有改变的株高的竹子突变体。这证明了CRISPR / Cas9在竹子中的适用性，从而促进了未来竹子的研究和育种。