CRISPR/Cas9‐mediated knockout of Ms1 enables the rapid generation of male‐sterile hexaploid wheat lines for use in hybrid seed production

Summary The development and adoption of hybrid seed technology have led to dramatic increases in agricultural productivity. However, it has been a challenge to develop a commercially viable platform for the production of hybrid wheat (Triticum aestivum) seed due to wheat's strong inbreeding habit. Recently, a novel platform for commercial hybrid seed production was described. This hybridization platform utilizes nuclear male sterility to force outcrossing and has been applied to maize and rice. With the recent molecular identification of the wheat male fertility gene Ms1, it is now possible to extend the use of this novel hybridization platform to wheat. In this report, we used the CRISPR/Cas9 system to generate heritable, targeted mutations in Ms1. The introduction of biallelic frameshift mutations into Ms1 resulted in complete male sterility in wheat cultivars Fielder and Gladius, and several of the selected male‐sterile lines were potentially non‐transgenic. Our study demonstrates the utility of the CRISPR/Cas9 system for the rapid generation of male sterility in commercial wheat cultivars. This represents an important step towards capturing heterosis to improve wheat yields, through the production and use of hybrid seed on an industrial scale.


Introduction
Global demand for food crops is projected to double between 2005 and 2050 (Tilman et al., 2011). In order to meet future demand and limit the environmental impact associated with doing so, new breeding technologies must be developed to increase crop yields (Tester and Langridge, 2010). Capturing heterosis through hybrid breeding is one of few crop improvement technologies that offers rapid and significant yield gains across diverse production environments. Hybrid seed has long been widely used for the production of major cereal crops such as maize and rice, but it has been a challenge to develop a commercially viable hybridization platform for bread wheat (Triticum aestivum) due to wheat's strong inbreeding features, and the absence of a simple, inexpensive and reliable system for hybrid seed production. Heterotic yield gains of more than 10% and enhanced yield stability have been observed in experimental wheat hybrids (Longin et al., 2013;M€ uhleisen et al., 2014), underscoring the potential of this breeding method. Given that wheat provides approximately one-fifth of dietary calories and protein for the human population (Shiferaw et al., 2013), it is clear that the development of a viable wheat hybridization platform could have a substantial positive impact on global food security.
Commercial hybrid seed production requires efficient crosspollination between genetically distinct parental inbred lines. In addition, self-pollination of the female parent must be avoided. In wheat, this has been difficult to achieve on a large scale due to a lack of efficient and reliable methods for separating the sexes and forcing outcrossing (Whitford et al., 2013). Recently, a novel hybridization platform that utilizes nuclear male sterility was described for maize (Wu et al., 2016). A key component of the platform is seed production technology (SPT), a process that enables the propagation of non-transgenic nuclear male-sterile inbred lines for use as female parents. This hybridization platform has since been extended to rice, a development that was made possible by the identification and isolation of the rice male fertility gene OsNP1 (Chang et al., 2016). Recently, we identified the wheat male fertility gene Ms1 by map-based cloning and demonstrated its function via complementation of the EMSderived mutation ms1d (Tucker et al., 2017). In our previous report, we also described how Ms1 could be used to establish SPT in wheat, and we highlighted the potential of genome editing for rapidly introducing highly penetrant recessive ms1 alleles into elite wheat lines.
The CRISPR (clustered regularly interspaced short palindromic repeats)/Cas9 (CRISPR-associated protein 9) system is currently the most widely used genome editing technology, largely due to its simplicity and flexibility. The system has two components that together form a ribonucleoprotein complex: the Cas9 endonuclease and a small guide RNA (gRNA). The gRNA contains a 20 nucleotide guide sequence that is designed to target a specific site (protospacer) in the genome via Watson-Crick base pairing. The protospacer must be located immediately 5 0 to a protospacer adjacent motif (PAM), whose canonical form is 5 0 -NGG-3 0 (Jinek et al., 2012). Following target site recognition, Cas9 creates a DNA double-strand break (DSB). Repair of the DSB through the error-prone non-homologous end-joining (NHEJ) pathway often leaves a lesion in the form of a small insertion/deletion (indel) mutation. Such mutations can shift the open reading frame of a coding sequence or introduce a pre-mature stop codon, resulting in gene knockout.
Here, we used the CRISPR/Cas9 system to generate Ms1 knockout wheat lines that exhibit male sterility in the first generation. One of the recessive ms1 alleles was highly penetrant and stably transmitted to the T 1 , T 2 and T 3 generations. Our results demonstrate the utility of the CRISPR/Cas9 system for the rapid generation of nuclear male sterility in hexaploid wheat. We anticipate that this approach will facilitate the development of a commercially viable wheat hybridization platform.
Inheritance of targeted mutations and male-sterile phenotypes in the T 1 , T 2 and T 3 generations To determine whether the targeted mutations were heritable, we tracked the inheritance of the +1 mutant allele in the progeny of the partially male-sterile line GL353-119 (+1/À3). As the +1 insertion of an A nucleotide created a new AluI restriction site, we were able to easily detect the mutant allele via AluI restriction enzyme assay.
By crossing GL353-119 (+1/À3) with wild-type cv. Gladius (Figure 4a), we obtained T 1 progeny (+1/WT or À3/WT), all of which were fully fertile ( Figure 4b). Line T1-1 (+1/WT) was selfed to produce 94 T 2 seeds that were either DsRed-positive (63 seeds) or DsRed-negative (31 seeds), based on fluorescence microscopy ( Figure 4c). The Cas9 transgene was detected by PCR in 59 of the 63 DsRed-positive T 2 plants (94%), while it was detected in only 5 of the 31 DsRed-negative T 2 plants (16%). The 26 DsRed/Cas9negative T 2 plants were genotyped for the +1 allele and phenotyped for fertility. All of the WT/WT and +1/WT plants were fully fertile, whereas all of the +1/+1 plants were malesterile and produced no seed ( Figure 4d and Table 2). The recessive +1 allele was inherited in a Mendelian fashion ( Figure 4e and Table 2). Thus, we obtained four male-sterile T 2 plants that were apparently non-transgenic (Table 2). Line T2-14 (+1/WT) was randomly selected and selfed to produce T 3 progeny. Fifty T 3 plants were genotyped and phenotyped. All of the WT/WT and +1/WT plants were fully fertile, whereas all of the +1/+1 plants were male-sterile, and the targeted mutations were inherited in a Mendelian fashion (Table 3). Additionally, the partially malesterile GL353-119 line produced a T 1 seed through selfing. The T 1 plant grown from this seed had the parental genotype (+1/À3), and it too was partially male-sterile.

Discussion
The CRISPR/Cas9 system is a powerful tool for studying gene function and generating genetic diversity in crops. In this study, our aim was to demonstrate the utility of the CRISPR/Cas9 system for the rapid generation of nuclear male sterility in hexaploid wheat. We hypothesized that knockout of Ms1 through the introduction of targeted biallelic frameshift mutations would result in male sterility, a trait that has high agronomic value for hybrid seed production.
Five per cent (2 of 40) of T 0 plants carrying gRNA LTPG1-2 were biallelic mutants, while the majority of the plants (26 of 40) were chimeras in which only a small proportion of cells were edited. Similar editing efficiencies have been reported for gRNAs targeting other wheat genes such as TaMLO-A1  and TaGW2 (Zhang et al., 2016). We observed substantial   [1905][1906][1907][1908][1909][1910][1911][1912][1913] variation in editing efficiencies between different gRNAs; of the three gRNAs tested, only one (LTPG1-2) had sufficient activity to generate heritable targeted mutations in Ms1. In silico prediction of gRNA activity was carried out using the sgRNA Designer (Doench et al., 2016) and WU-CRISPR (Wong et al., 2015) tools. These tools use different algorithms to calculate an activity score for each gRNA, where higher scores indicate higher predicted gRNA activity. WU-CRISPR has a score range of 0-100, but only gRNAs with scores ranging from 50 to 100 are displayed in the database as they are considered 'good' candidates (Xiaowei Wang, personal communication). LTPG1-2 was predicted by WU-CRISPR to be a good candidate (activity score = 70), whereas LTPG1-1 and LTPG1-4 (activity scores < 50) were not. Our experimental results were in agreement with these predictions. sgRNA Designer has a score range of 0-1, and there is no threshold for what is considered a 'good' candidate. LTPG1-1 (activity score = 0.4) was predicted by sgRNA Designer to be the most active of the three gRNAs, but our experimental results were not in agreement with this prediction. The lack of correlation between prediction and experiment in this case is not surprising, as we tested only three gRNAs. Given the variability in gRNA efficacy, and considering the time-consuming and laborious nature of wheat transformation and tissue culture, it is recommended that gRNAs be validated through transient expression in wheat protoplasts prior to commencing plant transformation .
We did not observe any off-target editing of Ms-A1 or Ms-D1 in T 0 lines. In the case of LTPG1-2, the presence of multiple mismatches (including at least one mismatch in the critical PAMproximal 'seed' region of the protospacer) was sufficient to abolish gRNA activity, as expected (Hsu et al., 2013;O'Geen et al., 2015;Sternberg et al., 2014;Wu et al., 2014). Thus, LTPG1-2 shows high specificity for Ms1 on chromosome 4BS. In the case of LTPG1-1 and LTPG1-4, on-target activity was very low or undetectable, and therefore, the lack of off-target activity was also expected.
We used three different methods for detecting targeted mutations in transgenic T 0 plants: capillary separation of fluorescently labelled amplicons, TIDE analysis of Sanger sequence traces and CRISPResso analysis of NGS reads. Only NGS had sufficient sensitivity to detect the low-frequency targeted mutations in the chimeras, but the three methods produced very similar results for the two biallelic mutants. Therefore, capillary separation of fluorescently labelled homoeolog-specific amplicons and TIDE analysis of homoeolog-specific Sanger sequence traces are rapid, reliable and cost-effective options for initial screening of T 0 plants for targeted mutations that are likely to be heritable.
As expected, biallelic knockout of Ms1 in the T 0 line FL353-19 (+1/+1) resulted in complete male sterility. By contrast, incomplete knockout of Ms1 in the biallelic mutant T 0 line GL353-119 (+1/À3) resulted in partial male sterility. The targeted mutations carried by GL353-119 were inherited in a Mendelian fashion, and completely male-sterile (+1/+1) mutants were recovered in the T 2 and T 3 generations, along with fully fertile +1/WT and WT/WT   plants. These results are consistent with previous reports of male sterility in EMS-derived Ms1 knockout mutants (Tucker et al., 2017;Wang et al., 2017). Furthermore, we obtained four T 2 knockout mutants that were apparently non-transgenic based on fluorescence microscopy and PCR assays. The identification of these mutants was streamlined by employing a fluorescencebased seed sorting strategy, similar to that which has been developed for Arabidopsis (Gao et al., 2016). Interestingly, the partial male-sterile phenotype, observed in GL353-119, was inherited by a +1/À3 T 1 plant. This suggests that the Leu7 residue (deleted in the À3 allele) is required for proper functioning of the signal peptide in Ms1. Genome editing has been successfully applied for the generation of male-sterile rice and sorghum lines (Chang et al., 2016;Cigan et al., 2017;Li et al., 2016;Zhou et al., 2016). Initial attempts to use (meganuclease-based) genome editing for the generation of male-sterile wheat lines  were met with limited success, as editing of only one of the three Ms26 homoeologs (A, B or D-genome) did not confer male sterility due to functional redundancy of the gene. However, conventional crossing of single-genome biallelic Ms26 mutants was successfully carried out to produce triple biallelic Ms26 mutants exhibiting male sterility . More recently, the CRISPR/Cas9 system was used to edit the wheat male fertility gene Ms45, and by selfing a triple monoallelic mutant T 1 plant, male-sterile triple biallelic mutant T 2 plants were recovered (Singh et al., 2018). Thus, gene functional redundancy can slow the process of recovering edited wheat lines with the desired phenotype, as the targeted mutations often need to be combined and/or made homozygous via conventional breeding.
Ms1 is a single copy gene located on chromosome 4BS (Tucker et al., 2017), and the homoeologs Ms-A1 and Ms-D1 are epigenetically silenced (Wang et al., 2017). This lack of functional redundancy among homoeologs makes Ms1 a particularly attractive target for genome editing. Indeed, our results demonstrate that the CRISPR/Cas9 system can be used to generate Ms1 knockout wheat lines that exhibit male sterility in the first generation. We also report here the sequence for LTPG1-2, an active gRNA that specifically targets Ms1 in cv. Fielder and cv. Gladius. Further studies will be needed to determine the efficacy of LTPG1-2 in other wheat cultivars, including elite germplasm from different breeding pools. If LTPG1-2 is found to be ineffective in a particular target cultivar, for example due to gene functional redundancy, then a different gRNA or multiple gRNAs may be needed. Furthermore, as we only tested three gRNAs, it is likely that screening of additional gRNAs would lead to the identification of a gRNA(s) that exhibits even higher activity than that observed for LTPG1-2.
Our study provides a methodological foundation and molecular tools for the rapid development of nuclear male-sterile wheat lines. This represents a significant step toward the establishment of a commercially viable hybrid wheat platform. Thus, we anticipate that the adoption of genome editing technologies for precision wheat breeding, together with a better understanding of wheat floral architecture and the flower opening process (Okada et al., 2018), will ultimately lead to increased yield gains through capturing heterosis.

Experimental procedures gRNA design
Partial Ms1, Ms-A1 and Ms-D1 sequences derived from T. aestivum cultivars Fielder and Gladius were used for gRNA design. The gRNAs were designed to target exon 1 of Ms1 on chromosome 4BS, in a region immediately downstream of the start codon, based on the presence of the canonical PAM (5 0 -NGG-3 0 ). Guide sequences were 20 nucleotides in length. To ensure efficient transcription from the TaU6 promoter, all gRNAs had a G nucleotide at position +20 (PAM-distal end) of the guide sequence (Sander and Joung, 2014).

Vector design and construction
All vectors were designed using Vector NTI software. The gRNA expression cassette (Shan et al., 2013) consisting of the TaU6 promoter and a non-targeting (random guide sequence) gRNA was synthesized (GenScript) and cloned into pUC57, resulting in pUC57-gRNA. Annealed oligos containing the Ms1-targeting guide sequence (Table S2) were cloned into pUC57-gRNA by simultaneous digestion/ligation with BbsI and T4 DNA ligase. Positive clones were identified by diagnostic restriction digest and validated by Sanger sequencing (Australian Genome Research Facility).
The rice codon-optimized SpCas9 gene with N-and C-terminal nuclear localization signals (Shan et al., 2013) was synthesized (GenScript) and inserted into PHP62407M as an NcoI-AscI fragment between the maize Ubi1 promoter and the Sorghum bicolor actin terminator, resulting in the entry vector pCas9-NB.
The Agrobacterium T-DNA binary vector pMDC123 (Curtis and Grossniklaus, 2003) was modified by replacing the original selection cassette with an intron-containing bar gene regulated by the maize Ubi1 promoter and the wheat rbcS Class II terminator. In addition, an aleurone-specific fluorescent reporter (DsRed2) cassette (Wu et al., 2016) was inserted between the Gateway cassette and the right border, resulting in the destination vector pMDC-Bar-DsRed. The Cas9 expression cassette from pCas9-NB was Gateway cloned into pMDC-Bar-DsRed to produce the intermediate vector pNB1.

Agrobacterium-mediated transformation
Transformation of cv. Fielder and cv. Gladius was carried out as described (Ishida et al., 2015), with minor modifications. Briefly, immature embryos were isolated from spikes harvested at 14 days post-anthesis. Isolated embryos were transferred to WLS-liq solution, centrifuged at 16 000 g for 10 min, incubated in WLS-inf solution containing Agrobacterium (strain AGL1) for 5 min and then transferred to WLS-AS media for 2 days of cocultivation. After co-cultivation, embryo axes were removed, and then, scutella were transferred to WLS-Res media for 5 days of resting culture. After the resting culture, scutella were transferred to WLS-P5 callus induction media (selection with 5 mg/L phosphinothricin) for 2 weeks, followed by WLS-P10 callus induction media (selection with 10 mg/L phosphinothricin) for 3 weeks. Calli were then transferred to LSZ-P5 regeneration media (selection with 5 mg/L phosphinothricin) for 2 weeks under a cycle of 12 h dark/12 h light (~70 lmol/m 2 /s). Regenerants were transferred to LSF-P5 rooting media (selection with 10 mg/L phosphinothricin) for 2 weeks, before being transferred to potted soil in the greenhouse. Timentin was substituted for cefotaxime in all tissue culture media.

Detection of targeted mutations by capillary separation of fluorescently labelled amplicons
Genomic DNA was extracted from the second leaves of transgenic T 0 plants at the vegetative stage, using a freezedried method (Kovalchuk, 2014). The target site was amplified by PCR using Phusion High-Fidelity DNA Polymerase (New England BioLabs), Phusion GC Buffer, 5% DMSO, 1 M betaine and a pair of Ms1-specific 6-FAM-labelled primers (Table S2).
To generate wild-type amplicons for spike-in (size reference), the same primer pair labelled with HEX was used. Touchdown PCR cycling conditions were as follows: initial denaturation at 98°C for 3 min, denaturation at 98°C for 15 s, annealing at 70-65°C for 20 s, extension at 72°C for 15 s and final extension at 72°C for 5 min. The starting annealing temperature was decreased by 0.5°C each cycle for 10 cycles, followed by 25 cycles at the final annealing temperature. A sample of the PCR product was run on an agarose gel to confirm the presence of a single band of the expected size (432 bp). The fluorescently labelled amplicons were diluted and subjected to capillary electrophoresis (Australian Genome Research Facility) on an AB3730 DNA Analyzer (Applied Biosystems, Foster City, CA). The results were analysed using PeakScanner Software 2 (Applied Biosystems). In PeakScanner, the peak for the HEX-labelled wild-type size reference was adjusted to a fluorescence intensity of approximately 3000, to improve visual clarity.

Detection of targeted mutations by Sanger sequencing and TIDE
Genomic DNA was extracted from transgenic T 0 plants as described above. The target site was amplified by PCR using Phusion High-Fidelity DNA Polymerase, Phusion GC Buffer, 5% DMSO, 1 M betaine and a pair of Ms1-specific primers (Table S2). Touchdown PCR cycling conditions were as follows: initial denaturation at 95°C for 8 min, denaturation at 94°C for 10 s, annealing at 62-57°C for 30 s, extension at 72°C for 30 s and final extension at 72°C for 5 min. The starting annealing temperature was decreased by 0.5°C each cycle for 10 cycles, followed by 30 cycles at the final annealing temperature. A sample of the PCR product was run on an agarose gel to confirm the presence of a single band of the expected size (577 bp). The amplicons were then column-purified and Sanger sequenced (Australian Genome Research Facility) on a 3730xl DNA Analyzer (Applied Biosystems). Bases were called with KB Basecaller v1.4.1.8, and the AB1 files were uploaded to the online TIDE analysis tool (Brinkman et al., 2014). In TIDE, the indel size range was set at 10, and the other settings were adjusted based on information provided on the online TIDE analysis tool Troubleshooting webpage. The proportion of edited DNA in the sampled tissue was calculated as the sum of all significant indels (P < 0.001) detected by TIDE.

Detection of targeted mutations by NGS and CRISPResso analysis
Genomic DNA was extracted from transgenic T 0 plants as described above. To generate amplicons for NGS, two rounds of PCR were carried out. The PCR mixtures contained Phusion High-Fidelity DNA Polymerase, Phusion GC Buffer, 5% DMSO and 1 M betaine. In the first round of PCR, the target site was amplified using conserved primers flanked by 5 0 universal tail sequences (Table S2). PCR cycling conditions were as follows: initial denaturation at 98°C for 3 min, followed by 30 cycles of 98°C for 20 s, 72°C for 20 s, with a final extension at 72°C for 2 min. A sample of the PCR product was run on an agarose gel to confirm the presence of bands of the expected sizes (300 bp, 313 bp, 327 bp). In the second round of PCR, barcodes and adapters were added using Illumina Nextera XT primers that anneal to the tail sequences of the primers used in the first round of PCR. Cycling conditions in the second round of PCR were as follows: initial denaturation at 98°C for 3 min, followed by six cycles of 98°C for 30 s, 55°C for 30 s, 72°C for 30 s, with a final extension of 72°C for 5 min. The barcoded PCR products were purified using Agencourt AMPure XP beads (Beckman Coulter), quantified by qPCR, pooled in equimolar amounts, spiked with 10% PhiX Control v3 and then sequenced (Australian Genome Research Facility) on the Illumina MiSeq platform using the MiSeq Reagent Kit v3 300 cycle. The raw reads from each sample were filtered using Ms1, Ms-A1 and Ms-D1-specific tag sequences (Table S2) and assigned to their respective homoeologs. FASTQ files containing the homoeolog-specific reads were used as input for the CRISPResso analyses (Pinello et al., 2016). In CRISPResso, the following parameters were used: -w 20 -hide_-mutations_outside_window_NHEJ -save_also_png -trim_sequences -q 30 -exclude_bp_from_left 5 -exclude_bp_from_right 5 -ignore_substitutions. Allele frequencies in Figure 3 were calculated by summing the values in the %Reads column of the CRISPResso allele frequency table, after filtering out aligned sequences that did not contain the partial allele sequence shown. Editing frequencies in Table 1 were calculated using data from the CRISPResso pie charts.

Pollen viability assay
Pollen viability was assessed by Lugol (1% I 3 K solution) staining. Pollen grains were mounted on glass microscope slides and imaged using a Nikon Ni-E microscope equipped with a DS-Ri1-U3 camera (Adelaide Microscopy Waite Facility). Images were captured with NIS-Elements software. AluI assay for genotyping T 2 and T 3 plants Genomic DNA was extracted from T 2 and T 3 plants as described above. The target site was amplified by PCR using Phusion High-Fidelity DNA Polymerase, Phusion GC Buffer, 5% DMSO, 1 M betaine and pair of Ms1-specific primers (Table S2). Touchdown PCR cycling conditions were as follows: initial denaturation at 95°C for 8 min, denaturation at 94°C for 10 s, annealing at 70-65°C for 30 s, extension at 72°C for 20 s and final extension at 72°C for 5 min. The starting annealing temperature was decreased by 0.5°C each cycle for 10 cycles, followed by 27 cycles at the final annealing temperature. A sample of the PCR product was run on an agarose gel to confirm the presence of a single band of the expected size (432 bp). Five lL of unpurified amplicons was digested with 2 units of AluI in a 7 lL reaction and then run on a 2% agarose gel.

Statistics
To test for Mendelian inheritance of edited alleles in the T 2 and T 3 generations, Pearson's chi-squared test was used, as described (Montoliu, 2012).
Post hoc in silico prediction of gRNA activity gRNA on-target activity was predicted using the sgRNA Designer (Doench et al., 2016) and WU-CRISPR (Wong et al., 2015) tools, according to the developers' guidelines.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Table S1 Transgene copy numbers for T 0 lines.