Using CRISPR/Cas9 genome editing in tomato to create a gibberellin‐responsive dominant dwarf DELLA allele

Summary The tomato PROCERA gene encodes a DELLA protein, and loss‐of‐function mutations derepress growth. We used CRISPR/Cas9 and a single guide RNAs (sgRNA) to target mutations to the PROCERA DELLA domain, and recovered several loss‐of‐function mutations and a dominant dwarf mutation that carries a deletion of one amino acid in the DELLA domain. This is the first report of a dominant dwarf PROCERA allele. This allele retains partial responsiveness to exogenously applied gibberellin. Heterozygotes show an intermediate phenotype at the seedling stage, but adult heterozygotes are as dwarfed as homozygotes.


Introduction
Genome editing methods have great promise for functional genomics research and crop improvement. The CRISPR (Clustered Regularly Interspaced Short Palindromic repeat)/Cas9 system is a powerful tool for genome editing. When this prokaryotic system is engineered for use in eukaryotes, a short RNA molecule, often expressed from an RNA polymerase III-dependent promoter such as that of the U6 splicing factor RNA, guides the associated endonuclease Cas9 to generate double strand breaks 3 base pairs (bp) 5 0 to the protospacer adjacent motif (PAM 5 0 -NGG-3 0 ) in the target genomic DNA (Jinek et al., 2013;Ran et al., 2013;Sinkunas et al., 2013). Inaccurate repair of this break by nonhomologous end joining (NHEJ) within an exon often produces insertion/deletion (Indel) mutations that cause translational frame shifts or amino acid replacements or deletions.
The first reports of CRISPR/Cas9 editing in plants appeared in 2013. CRISPR/Cas9 editing has been reported in the model plants Arabidopsis thaliana and Nicotiana benthamiana, in rice, sorghum, wheat, tomato, grapes, maize and the opium poppy (Alagoz et al., 2016;Liu et al., 2017). CRISPR/Cas9 has now become the tool of choice for gene editing in plants not only to knock out gene(s) but also to insert or delete a gene (Filler Hayut et al., 2017;Roux et al., 2006) by using CRISPR/Cas9-mediated DNA breaks to elevate levels of Homologous Recombination (Cermak et al., 2017;Cerm ak et al., 2015).
Plant growth is promoted by gibberellin, a hormone that is involved in diverse developmental processes, including stem elongation, leaf expansion, pollen development, flowering and seed germination. Gibberellins are diterpenoid acids that are synthesized by the terpenoid pathway in plastids and then modified further in the endoplasmic reticulum and cytosol to create the active form. The gibberellins are named GA 1 through GA n in order of discovery. GA3 was the first gibberellin to be identified, and although not the major active GA in most plants, was used in our experiments (Eriksson, 2006;Olszewski et al., 2002). Altering GA signalling disrupts control of cell and organ size (Fleet and Sun, 2005). GA promotes degradation of DELLA proteins, which are nuclear-localized, negative growth regulators (Itoh et al., 2002;Olszewski et al., 2002;Wen and Chang, 2002). GA binding to the soluble Gibberellin Insensitive Dwarf (GID) receptor triggers GID interaction with the DELLA proteins, which then stimulates the assembly of an SCF E3 ubiquitin ligase complex that contains the GID2/SLEEPY1 F-box proteins (Ueguchi-Tanaka et al., 2005). This SCF complex polyubiquitinates the DELLA proteins, leading to their degradation by the 26S proteasome, thus derepressing growth (Livne et al., 2015). The A. thaliana genome encodes five highly homologous DELLA proteins including GA-insensitive (GAI) and repressor of ga1-3 (RGA) (Peng et al., 1997). A 17-amino acid deletion affecting the conserved DELLA motif results in a dominant dwarfing allele of GAI. Other dominant dwarfing alleles of the DELLA repressors include reduced height1 (Rht1) from wheat (Triticum aestivum), dwarf8 (d8) from maize (Zea mays) (Willige et al., 2007) and Sln1D from Barley Gubler et al., 2002). These alleles encode DELLA proteins that retain the ability to repress growth but are not destabilized by GA due to impairment in binding to the GA-GID1 complex. In most cases these dwarfs are non-responsive to exogenous GA. Recessive loss-of-function DELLA (slender) mutants in barley, tomato and rice are highly elongated and male sterile (Asano et al., 2009;Chandler and Harding, 2013;Chandler et al., 2002;Ikeda et al., 2001). Arabidopsis and barley DELLA alleles have been found to uncouple the meristem and inflorescence size from plant height (Serrano-Mislata et al., 2017).
Tomato is an important crop with 223 million tonnes produced in 2014 (Food and Agriculture Organization of the United Nations Database, 2014). Significant resources are invested for breeding traits such as disease resistance, fruit shape and colour. Tomatoes with reduced plant height and compact growth habits could be useful in some environments. Most currently grown glasshouse cultivars tend to be tall, trailing plants that require additional mechanical support, pruning and side-shoot management (Hochmuth, 2001;Hochmuth et al., 1997). This study reports for the first time a dominant dwarf mutation in tomato using CRISPR technology, which may prove useful in the tomato industry.

Results and discussion
Editing the tomato DELLA-encoding PROCERA gene We generated constructs carrying a single guide RNA that targets the nucleotides encoding the amino acids DELLAVLG of the tomato DELLA gene (Figure 1a,b) (see Experimental procedures). The DELLA target sequence GGATGAGCTTTTAGCTGTTT was submitted to a Blast search against tomato database using http:// www.rgenome.net/cas-offinder/, and no other target was found. To detect mutations introduced by the Cas9 nuclease in T 0 plants, we assessed loss of a restriction enzyme site that might have arisen due to imprecise NHEJ repair (Belhaj et al., 2013(Belhaj et al., , 2015Nekrasov et al., 2013;Voytas, 2013) (Figure 2a-c). Genomic DNA (gDNA) was extracted from the 46 T 0 lines (individual transformants) (see Experimental procedures). For each of the T 0 plants we PCR-amplified the DELLA-encoding domain and then digested with AluI and examined products on a 3% agarose gel ( Figure S1). The AluI site is located 3 bp away from the Cas9 nuclease cutting site and small indels that extend into the AluI site should result in loss of AluI digestibility. We did not observe primary amplicons smaller than 114 bp, suggesting large deletions were not found in our experiments. Eight lines 4, 5, 6, 15, 23, 24, 27 and 40 showed a 114 bp non-digested band in addition to two fragments (51 and 63 bp) produced when AluI digests wild-type DNA, suggesting that the Cas9 nuclease was active in somatic tissues. Purified non-digested 114 bp PCR fragments from lines 4, 24, 27 and 40 were cloned into a TA cloning vector and sequenced. Analysis of the sequences revealed three types of mutations: either a 5 or a 3 nucleotides (nt) deletion, or a T insertion (Figure 2c). Of the 46 T 0 lines, only line 40 showed a strong AluI undigested band. Sequencing of the non-digested band reveals the presence of wild type (WT) and DELLV fragments (Figure 2c).
A heritable dwarf phenotype in progeny of T 0 plants T 0 plants were self-pollinated and T 1 progeny plants were analysed either by restriction site loss assay or phenotypic screening. The fruits from lines 4, 5, 6 and 15 did not produce any seeds. When progeny from line 40 were grown, five of the plants had a WT stature and eight of the plants were smaller ( Figure 3a) suggesting that an editing event might have occurred early during development (see also Figure 2c). Genomic DNA from these individuals was extracted and PCR with primers flanking the target (Primers 1 and 2) was performed. Restriction enzyme site loss analysis of these individuals revealed that three were homozygous mutants, five contained the AluI resistant band as well as the two WT bands and were scored as heterozygous, and five were WT ( (Figure 3c). In Arabidopsis, GA perception results in DELLA protein ubiquitination by SCF SLY/GID2 and then degradation by the proteasome, resulting in DELLA protein destruction that de-represses growth (Qin et al., 2014). Deletion of 17 amino acids containing the DELLA motif is responsible for the gain-of-function properties of the gai-1 allele which is severely dwarfed (Ariizumi et al., 2008). Several DELLA domain mutations have been described that result in GA-insensitive growth in different plant species including barley, Arabidopsis, wheat and rice. Dominant dwarf DELLA protein forms are refractory to degradation upon GA perception. PRO D /PRO tomato plants have one copy of the WT allele, and one copy of the DELLA mutant form that is expected to be refractory to degradation. Failure to degrade the DELLVLG protein causes growth retardation. It was also reported that there is a correlation between the severity of the dwarf phenotype and the stabilization of the DELLA protein caused by the mutation (Willige et al., 2007).
In the fragment we chose to amplify, a second AluI recognition site is present 13 bp away from the PAM. PCR primers that removed the AluI site in the forward primer were designed to leave intact the second AluI site closer to the PAM, where it is likely to be altered by imperfect NHEJ. Two primers flanking the DELLA target locus were designed to amplify a fragment of 114 bp, digested with AluI and then visualized on a 3% agarose gel ( Figure 5). PRO D /PRO D mutants are smaller than WT plants at 6, 9 and 11 weeks old ( Figure 5). The diameter of PRO D /PRO D fruit was slightly but not significantly smaller than WT; however, fruit weight was slightly and significantly reduced (Data S1). The PRO D /PRO D fruits contain an average of 35 fewer (~57 versus 92) seeds per fruit compared to the WT ( Figure S2). The flowering time and fruit setting time were indistinguishable between the two genotypes ( Figure S3). The number of tomato fruit per plant was calculated and show that PRO D /PRO D and WT produce the same average number of fruits per plant.

PRO D /PRO D plants retain partial responsiveness upon GA treament
We investigated the GA response of PRO D /PRO D tomato lines. Mutant and WT plants were sprayed to run-off with 50 lM gibberellic acid (GA 3 ) (see Experimental procedures). The results showed that PRO D /PRO D plants, while growing slower than WT, retain partial responsiveness to GA treatment ( Figure 6). A similar observation was made on the barley Sln1D allele where a single amino acid exchange mutation DELLAVLG to DELLAVLE showed an intermediate phenotype with respect to plant growth and mutant GAI protein stabilization Gubler et al., 2002;Willige et al., 2007). A dwarf maize mutant was also reported carrying a single amino acid insertion in a domain located downstream of the DELLA domain, called VHYNP domain of the dwarf8 gene and shows a strongly reduced but not abolished response to GA (Cassani et al., 2009). Further studies with the Arabidopsis GID1A GA receptor and the GAI DELLA repressor mutant proteins indicate that the loss (or reduction in the case of Sln1D) of GA responsiveness can be explained by the loss (or reduction) of GA-dependent GID1A binding. Thus, our observations are consistent with other reports in the literature that partial interference with DELLA degradation by small deletions in the DELLA domain result in a weak or partially dominant dwarf phenotype Gubler et al., 2002;Willige et al., 2007).

Recovery of PROCERA loss-of-function mutants in the progeny of T 0 plants
Several lines 1, 13, 23, 25, 33 and 46 that showed no alteration in AluI restriction pattern (2 bands of 51 and 63 bp) were sown alongside WT plants (Figure 7a). When five, 2 and 72 individuals from lines 13, 23 and 25 were grown, one, one and three plants respectively in these populations were taller than their siblings. Similar to what has been observed in PROCERA mutants (Bassel et al., 2008;Jasinski et al., 2008), the leaves of these individuals bear fewer leaflets and the leaflets have a smoother margin (Figure 7b,c). Genomic DNA from these individuals was extracted, PCR-amplified with primers flanking the DELLAencoding domain (Primer 1 and 2) and PCR products cloned into T/A cloning vectors. Sequencing analysis revealed a T insertion next to the PAM motif leading to an early stop codon just after the DELLA motif for all of them (-MDELLAVFGL*) ( Figure 7d), such mutations would not be revealed by the AluI digestion assay since the AluI site is located 3 bp away from the Cas9 nuclease cutting site. The previously reported PROCERA mutant phenotype was due to a T905A mutation leading to the conversion of VHVID to VHEID in the C-terminal GRAS motif, which is thought to be important for DELLA action but has no defined biochemical function. A PRO mutant allele was obtained using TALENs in order to determine the role of PROCERA in GA signalling (Livne et al., 2015). In contrast with the first characterized PROCERA allele which carries a missense mutation and shows partial responsiveness to GA, isolated proTALEN and proΔGRAS mutations are complete null alleles and GA insensitive (Livne et al., 2015). The targeting of the DELLA motif using CRISPR Cas9 in our experiments thus resulted in the generation of two phenotypes, one (PRO D /PRO D ) showing a dwarf phenotype and the second a PROCERA-like phenotype caused by loss of PROCERA function.

Conclusion
CRISPR technology is a powerful tool for genome editing that enables modification of gene(s) of interest, and since the target DNA is unlikely to be linked to the T-DNA, the Cas9 T-DNA locus can be segregated away by crossing. We report here the first use of CRISPR technology to engineer a dominant gain-of-function mutation that potentially could confer a desirable commercial trait. Over the past decades, DELLA gene mutations have been used by plant breeders. The most notable example is the dominant semi-dwarf wheat cultivars, a crucial component of the 'green revolution' genes which all contain deletions in the DELLA region of an Rht gene (Pearce et al., 2011;Peng et al., 1999). Our report paves the way for facile provision of dominant semi-dwarf alleles in other crops.

Plant growth and transformation
Transformations were conducted using the tomato cultivar Moneymaker GCR758 (Solanum lycopersicon). The binary Level 2 vector pAGM4723 containing the cassette nos prom :nptII: ocs ter , 2x35S prom :hSpCas9:ocs ter and DELLA gRNA were mobilized into Agrobacterium tumefaciens Agl1. Transgenic plants were regenerated by standard methods for tomato transformation (Fillatti et al., 1987). All the independent transformed lines were analysed. Plants derived from tissue culture were rooted in Jiffy peat plugs. Plants were potted into 7 cm pots using Levington M2 peat based compost and placed in a closed unheated propagator in the glasshouse. After 24 h in the propagator, they were vented (approx. 5%) and the resilience of the plants to external glasshouse conditions was tested. Plants were gradually acclimatized to the glasshouse conditions and then grown in Levington's M3 compost (Levington Horticulture Ltd., Fisons, Ipswich, UK) in a greenhouse at temperatures between 22 and 27°C in the light and 12-16°C in the dark. Sixteen hours of light was supplied at a photon flux density of 300-650 lE/m 2 /s at the plant surface, and the relative humidity was between 70% and 80%.

Golden Gate Plasmid constructs
We used the methods for designing single guide RNAs (sgRNAs) described in Nekrasov et al. (2013) to target the DELLA motif. The DELLAVLG amino acid sequence of PROCERA (PRO) is encoded by GATGAGCTTTTAGCTGTTTTGGGT, and we chose GGATGAG CTTTTAGCTGTTT as the target sequence with TGG as the PAM. This sequence was fused to a single gRNA (Jinek et al., 2012). The 20nt DELLA gRNA was amplified with the following primers DELLA gRNA_F gaGAAGACaaATTggatgagcttttagctgtttGTTTTAG AGCTAGAAATAGC and DELLA gRNA_R gaGAAGACaaAGCG TAATGCCAACTTTGTAC using as a template the Addgene Plasmid # 46966. The 158 bp resulting fragment was gel purified using a Qiagen gel extraction kit and subsequently cloned into the L0 module using Golden Gate technology (Engler et al., 2009(Engler et al., , 2014Patron, 2014). RNA Pol III transcribes U6 RNA genes of Arabidopsis, and a functional U6 promoter contains an Upstream Sequence Element and TATA-like elements. We engineered a U6 synthetic promoter which is the consensus sequence of the three U6 promoter variants identified in the Arabidopsis genome Waibel and Filipowicz, 1990). The resulting Level 2 binary vector (L2) in pAGM4723 carries a kanamycin resistance cassette in position 1 (Addgene Plasmid # 51144), 2x35S prom :hSpCas9:ocs ter in position 2 (Addgene plasmid # 49771) and U6: DELLA gRNA in position 3 ( Figure 1b) (Engler et al., 2014;Patron, 2014). BsaI-HF and BbsI from New England Biolabs and Thermo Fisher Scientific respectively were also used for the building up of the Golden Gate constructs.

T/A cloning and Sanger sequencing
We used the pGEM-T kit from Promega, Southampton, United Kingdom for TA cloning of PCR products. The Sanger sequencing of the constructs was done by GATC Biotech.

Genomic DNA extraction, restriction loss assay and PCR assays
The majority of DNA extraction methods from plant leaf tissue are derived from the original hexadecyltrimethylammonium bromide (CTAB)-based method described by (Doyle and Doyle, 1987). One hundred nanograms of DNA from each individual was used to perform a PCR with Primer1 GGGAAGAAGATGAAGAAGA AAAACCAGATGCTGGAATGGATGCGCTT and Primer2 GCT CAAAAACTTGAACAGC. Oligonucleotides were purchased from Merck, Gillingham, Dorset, Great Britain. PCR amplifications were conducted using New England Biolabs, Wilbury Way, Hitchin, Hertfordshire, Taq DNA polymerase (Catalogue: M0273S) and following the manufacturer protocol with a 25 lL final volume. Mixtures were amplified in a Bio-Rad C1000 Touch PCR machine Watford, Hertfordshire, UK. The thermal cycle was programmed for 30 s at 95°C as initial denaturation, followed by 35 cycles of 30 s at 95°C for denaturation, 30 s at 55°C as annealing, 15 s at 72°C for extension, and final extension at 72°C for 5 min. Five microliters of the PCR products were examined by electrophoresis at 100 V for 30 min in a 3% (w/v) agarose gel in 19 TBE buffer containing ethidium bromide. The gel was then observed under UV light. Restriction loss assay was conducted according to (Belhaj et al., 2013) using AluI. One unit per reaction of AluI from New England Biolabs was used for the restriction digest. PCRs to detect the presence of Cas9 gene were conducted in the same conditions as described above except the extension was 45 s and the PCR products were examined by electrophoresis in a 1.5% (w/v) agarose gel; the primers used were Cas9_Fwd: TTTCAGATTTCAGAAAGGACTTTCAG and Cas9_Rev: AGAtCCTTTGAGCTTTTCATAGTGGCTGG.

Gibberellic acid growth assays
Tomato seeds were planted in BM2 Germinating Mix (Berger Peat Moss, Saint-Modeste, Quebec, Canada) initially saturated with 1.5 g/L 20 : 10 : 20 20-10-20 Peat-Lite Fertilizer (JR Peters Inc, Allentown, PA). Tomato plants were grown for 4 weeks in flats spaced such that they would not significantly impede each other's growth. After 4 weeks, the tomato plants were measured and then treated by spraying to runoff with 50 lM GA 3 (Sigma, Gillingham, Dorset, United Kingdom; G7645; day 1 of treatment). As the GA 3 stock was dissolved in ethanol, the control tomato plants were treated with water containing the equivalent amount of ethanol. The GA 3 stock solution consisted of 50 mM GA 3 in 70% ethanol, and 1 mL/L was added to make the final 50 lM solution. In the control solution, 1 mL/L 70% ethanol was added. Plants were treated on days 1, 3 and 5. Measurements were taken on days 1, 3, 5, 7, 9, 11, 13 and 15.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article: Figure S1 Restriction site loss assay of the 46 T 0 tomato Moneymaker lines. Figure S2 Graphs showing the seed number per fruit, fruit diameter (mm), height (mm) and weight (gram) for PRO D /PRO D and WT.