Liyu Hung,Ru Zhng,Gungfu Hung,Ynxi Li,Gethew Melku,Shili Zhng,Hito Chen,Ynjun Zho,Jing Zhng,Yesheng Zhng,,*,Fengyi Hu,**

aState Key Laboratory for Conservation and Utilization of Bio-Resources in Yunnan/Research Center for Perennial Rice Engineering and Technology of Yunnan/School of Agriculture,Yunnan University,Kunming 650091,Yunnan,China

bState Key Laboratory of Genetic Resources and Evolution,Kunming Institute of Zoology,Chinese Academy of Sciences,Kunming 650223,Yunnan,China

cBGI-Baoshan,Baoshan 678004,Yunnan,China

Keywords:Rice CRISPR/Cas9 Gn1a DEP1 Gene editing

A B S T R A C T Rice yield is an important and complex agronomic trait controlled by multiple genes.In recent decades,dozens of yield-associated genes in rice have been cloned,many of which can increase production in the form of loss or degeneration of function. However,mutations occurring randomly under natural conditions have provided very limited genetic resources for yield increases.In this study,potentially yield-increasing alleles of two genes closely associated with yield were edited artificially.The recently developed CRISPR/Cas9 system was used to edit two yield genes:Grain number 1a(Gn1a)and DENSE AND ERECT PANICLE1(DEP1).Several mutants were identified by a target sequence analysis.Phenotypic analysis confirmed one mutant allele of Gn1a and three of DEP1 conferring yield superior to that conferred by other natural high-yield alleles.Our results demonstrate that favorable alleles of the Gnla and DEP1 genes,which are considered key factors in rice yield increases,could be developed by artificial mutagenesis using genome editing technology.

1.Introduction

Because rice is one of the most important crops and supports more than half of the world population[1],rice breeders target mainly improvement of its yield.Rice yield is a complex agronomic trait that is usually defined by number of panicles per plant,number of grains per panicle and thousand grain weight[2].Dozens of genes regulating these yield-associated traits in rice have been cloned(Table S1).Most of these structural genes,such as DEP1[3],Gn1a[4],GRAIN SIZE3(GS3)[5],GRAIN WIDTH5(GW5)[6],were negatively regulated yield and always be weaken for increasing yield.Under natural conditions, such desirable mutations in yield-associated genes are usually favored and consequently fixed by both farmers and breeders[3–6].Although these mutations have contributed substantially to rice yield increase, there are considerations of natural alleles having the best mutations.

Recently,genome-editing technologies have been successfully applied in a variety of plant species to target mutagenesis or to perform editing using zinc-finger nucleases(ZFNs)[7],transcription activator-like effector nucleases(TALENs)[8]and clustered regularly interspaced short palindromic repeats(CRISPR)-associated(Cas)systems(CRISPR/Cas9)[9].Owing to its technical simplicity and efficiency,the CRISPR/Cas9 system is widely used in advanced genome engineering and functional genomic studies of crops such as sorghum[10],rice[11],wheat[12], soybean [13],and maize [14]. In rice, genome-editing technologies have provided novel alleles of yield-associated genes[15].For the purpose of breeding,these alleles are more effective than the natural ones[15,16].Recently[17],four rice yield genes,Gn1a,DEP1,GS3,and IDEAL PLANT ARCHITECTURE1(IPA1)in a Zhonghua 11 background were edited for the first time using CRISPR/Cas9 technology.Shen et al.[18]subsequently targeted the disruption of GS3 and Gn1a in five widely cultivated japonica varieties using the CRISPR/Cas9 system and obtained multiple mutated alleles of both genes.These two studies demonstrated that the newly obtained mutations in different genetic backgrounds conferred the same yield increases as the natural mutations. However, neither study revealed a stronger effect of the alleles created by gene editing than found for the natural alleles.

DEP1 and Gn1a are two main yield-associated genes that can be subjected to mutation for developing superior alleles in rice[3,4].DEP1 is a well-known rice yield-associated gene encoding a phosphatidylethanolamine-binding protein (PEBP) [3]. Open reading frame(ORF)disruption by the replacement of a 637-base pair(bp)stretch of the fifth exon with a 12-bp sequence resulted in reduced length of the inflorescence internode,an increased number of grains per panicle and a consequent increase in grain yield as high as 40.9%in NIL-dep1 rice plants[3]. Gn1a is a gene for cytokinin oxidase/dehydrogenase(OsCKX2),in which deletions of 6 bp in the first exon,11 bp in the third exon,or 16 bp in the 5′-untranslated region increased yield by as much as 21%in NIL-Gn1a plants[4].

Phenotyping and selection for superior alleles can result in a significantly higher yield than that conferred by the natural alleles[19,20].Such superior alleles obtained through artificial mutagenesis can provide novel materials for the improvement of current rice varieties.To obtain high-yielding rice cultivars,artificial mutagenesis of yield-regulating genes via gene editing is a novel approach.This study accordingly aimed at mutating the two yield-regulating genes DEP1 and Gn1a in order to develop superior alleles in rice.

2.Materials and methods

2.1.Construction of a high-efficiency Agrobacterium-mediated CRISPR/Cas9 system

The CRISPR/Cas9 binary vector pCAMBIA1300-OsU3-Cas9 was used to construct target vectors of Gn1a and DEP1(Fig.1-A).The target sequences were introduced into the gRNA expression cassettes in the vector using two sites of the restriction enzyme Aar I.To use OsU3 as a promoter for sgRNA,an adenine was made the first nucleotide of the target sequence.In addition,GGCA and AAAC overhangs were added to the 5′termini of the paired complementary oligonucleotides.Target sequences of Gn1a and DEP1,which link in front of the sgRNA cassette,were designed in the fourth exon of Gn1a and the fifth exon of DEP1,respectively(Figs.2;S1).In particular,one target sequence that could recognize two adjacent sites for DEP1 was selected(Fig.S1-B).For OsPDS,DEP1,and Gn1a genes,oligonucleotides of target sequences were synthesized(Table S2). The CRISPR/Cas9 plasmid was digested to be linearized by the Aar I enzyme.The two complementary oligonucleotides were then annealed,phosphorylated,and ligated to the linear plasmid with T4 DNA ligase(Fig.1-A,B).The resulting ligation products were transformed into DH5a E.coli competent cells and coated on plates,and a monoclonal colony was selected for Sanger sequencing using OsU3-test-F/OsU3-test-R primers(Table S2).

2.2.Generation of transgenic rice

Transgenic rice was developed as described by Yukoh et al.[21].The two binary vectors pGn1a and pDEP1 were transformed into the Agrobacterium strain EHA105 by electroporation and then further transformed into Nipponbare,a japonica rice cultivar.

2.3.Genotyping mutations in rice

2.4.Phenotype observation of mutated lines

To obtain T2plants and evaluate the effects of the mutated alleles on rice yield,seeds of all 13 T1homozygous mutants and wild types selected from native negative T0lines were collected and grown in the field.Forty-two T2plants were grown in one paddy field for each allele and yield was measured.For the T2plants with the same genotype as T1,accurate assignment of each allele was ensured by double confirmation of the genotype of the corresponding target for each mutant.At maturity,yield per plant in both mutant and wild type was recorded.The whole plants were phenotyped for thousand grain weight,setting rate,panicle number,grain number,plant height,heading time,number of tillers,and plant architecture.The major panicle was further phenotyped for numbers of primary and secondary branches and grains number. A field yield test was performed, based on a randomized block design with three replications.All phenotyping experiments were performed at Jinghong,in the south of Yunnan province.

Fig.1–Flow diagram of artificial mutagenesis.A:pCAMBIA1300-OsU3-Cas9 binary vector,where the hSpCas9 cassette is driven by the 2×35S promoter and sgRNA(simple guide RNA)is controlled by the rice OsU3 promoter.The target region contains two Aar I enzyme sites,and HYG(hygromycin)was used as a resistance selection marker.B:Basic process of targeting genes for CRISPR/Cas9 constructs.C:Albino phenotype caused by OsPDS mutation.D:Gel image of the target gene after Pst I digestion.M,marker.1–9,mutation lines of OsPDS.wt,wild type.The mutation in CTGCAG(Pst I site)cannot be cut by the Pst I enzyme.E:Sequence of some mutation lines aligned with wild-type OsPDS gene sequence.?11 and+1 represent an 11-bp deletion and a 1-bp insertion,respectively.

3.Results

3.1.Artificial mutagenesis in OsPDS,Gn1a,and DEP1 genes

A binary vector (pCAMBIA1300-OsU3(Aar I)-Cas9) that contained the Cas9 cassette driven by the 35S promoter and simple guide RNA (sgRNA) cassette under the rice OsU3 promoter was constructed for efficient rice mutagenesis(Fig.1-A).The proper functioning of this CRISPR/Cas9 system was tested by the detection of an albino phenotype from a mutation in the phytoenedesaturase-encoding(OsPDS)gene[23].Besides counting albino phenotypes in T0generations(Fig.1-C),DNA sequencing and digestion of PCR products were also used for the detection of mutagenic efficiency(Fig.1-D,E).Fifty-four independently transformed rice lines were obtained,and eight lines showed the albino phenotype,including six biallelic and two homozygous mutations in the OsPDS gene. The CRISPR/Cas9 system resulted in a mutagenic efficiency of 28% (15/54), with 3.7% (2/54) homozygous mutations in the T0generation.

Transformed rice plants,independent lines defined as T0,were regenerated, including 53 individuals of the Gn1a genotype and 48 of the DEP1 genotype.The disrupted target sites detected by PCR-based Sanger sequencing revealed 49%(26/53)and 39%(19/48)targeting efficiency for Gn1a and DEP1,respectively(Table 1).Only 3 and 1 homozygous mutations were identified in the respective T0generations of the Gn1a and DEP1 transformed rice plants.

3.2.Genotyping of Gn1a and DEP1 artificial mutagenesis lines

In the T1generation grown from seeds of T0plants,homozygous mutants at the target sites and lacking the T-DNA segments were screened.Except for multiple mutations(alleles)in a single plant(Table S3),six and seven homozygous novel alleles in T1marker-free plants were obtained for Gn1a and DEP1,respectively(Figs.2;S1).All the artificial alleles were different from the natural ones,and most of them represented frameshift mutations(Fig.2).In more detail,two alleles were inserted(+1 bp)and four alleles were deleted(?1 bp,?4 bp,?21 bp,and ?1 bp)for Gn1a,and one allele was inserted(+1 bp)and six alleles were deleted(?1 bp,?6 bp,?18 bp,?2 bp,?21 bp,and ?1 bp)for DEP1(Figs.2;S1).These new mutant alleles were named Gn1a-G1 to Gn1a-G6 and DEP1-D1 to DEP1-D7(Figs.2;S1).Protein sequence comparison between wild-type and mutant plants revealed 33,32,and 33 amino acid(aa)extensions to the respective Gn1a-G1,Gn1a-G4,and Gn1a-G6 proteins.In contrast,the other three mutations in G2,G3,and G5 encoded truncated proteins,with respectively 46,18,and 7 aa reductions(Fig.2-A).

Except for DEP1-D7(+16 aa),truncated homologous mutants(DEP1-D1 to DEP1-D6 with ?2 to ?131 aa)were observed mainly in DEP1(Fig.2-B).Compared to the wild type,the ORFs of DEP1-D2(?131 aa)and DEP1-D5(?130 aa)were severely truncated(Fig.S1-B).The CRISPR/Cas9-engineered DNA nicks tended to be repaired to form deletion mutations,leading to more deletions than insertions in the target sites of both Gn1a and DEP1.

Fig.2–Sequencing results of homozygous Gn1a(A)and DEP1(B)mutant rice plants in the T1 generation.Sequences above the peak are Gn1a(A)and DEP1(B)wild-type alleles,respectively.The peak of each base for the whole sequence was single and revealed homozygosity.“+”shows insertion and“?”shows deletion,and PAM stands for the Protospacer Adjacent Motif.Wild indicates the wild-type Nipponbare.

3.3.Phenotyping yield-associated traits of Gn1a mutants

In the alleles created by artificial mutagenesis of Gn1a(G1 to G6),G2,G3,G4,and G6 alleles increased yield per plant,and three(G3,G4,and G6)increased the yield significantly by 21.1%,23.8%,and 11.4%, respectively (Fig. 3-E). Strikingly, the G4 allele increased yield per plant by 23.8%over the natural high-yield allele (21%) [4]. Comparison among the four main yieldassociated traits of both mutants and wild types indicated that grain number per panicle was a stronger contributing factor than setting rate,thousand grain weight,or panicle number per plant(Fig.3-A to D),in agreement with previous report[4].Counts of primary and secondary branches along with kernelcounts from the major and all other panicles revealed a strong association between the number of secondary branches and high grain number per plant(Figs.3;S2-A to C).Fig.4-A also shows that the field yields of G3 and G4 were significantly higher than those of the wild type.No other agronomic traits of any of the Gn1n mutants differed significantly from those of the wild type(Table S4).

Table 1–Targeting mutagenesis efficiency in T0 rice plants.

Fig.3–Phenotype analysis of the novel alleles of Gn1a and their comparison with the wild-type alleles.Setting rate(A),grain number per panicle(B),thousand-grain weight(C),panicle number per plant(D),and yield per plant(E)are displayed in histograms.Values are means±SD(n=24).*and**indicate significant differences from the wild type at P ＜0.05 and P ＜0.01,respectively,by Student's t-test.Panicle number is the number of effective tillers that were fertile.Panicle phenotypes(F)include primary and secondary branch numbers and grain number.Wild indicates the wild type selected from negative lines.

3.4.Phenotyping yield-associated traits of DEP1 mutants

Each of the seven artificial alleles of DEP1(D1 to D7)was sequenced and significantly increased yields per plant,by 13.0%,46.1%,50.7%,33.8%,51.1%,21.3%,and 24.7%,respectively(Fig.5-E).Three alleles(DEP1-D2,DEP1-D3,and DEP1-D5)conferring more yield increase than the natural allele(40.9%) [3] increased yield by 46.1%, 50.7%, and 51.1%,respectively.Thousand grain weight and grain number per panicle showed a slight decrease(Fig.5-B,C).Interestingly,panicle number per plant was the main contributor to higher kernels per rice plant as well as higher yield per plant in the DEP1 mutants(Figs.5-D,E;S-2-D to F;S3).For DEP1-D5,which increased rice yield by as much as 51.1%,the number of secondary branches of the average panicle had a significant contribution to the kernel yield and number of panicles per plant(Figs.5;S2-E,F;S3).Fig.4-B also shows that the field yields of D2,D3,D4,and D5 were significantly higher than those of the wild types.Like panicle number per plant, numbers of tillers of all the DPE1 mutants were significantly higher than that of the wild type. Other agronomic traits including plant height,heading time,and plant architecture of the DEP1 mutants did not differ significantly from those of the wild type(Table S4).

Fig.4–Field yield of plants carrying novel alleles of Gn1a(A)and DEP1(B),displayed in histograms.Values are means±SD(n=3).**indicates a significant difference from the wild type at P ＜0.01 by Student's t-test.Wild indicates the wild type selected from negative lines.

Fig.5–Phenotypes of the novel alleles of DEP1 and their comparison with the wild-type allele.Setting rate(A),grain number per panicle(B),thousand-grain weight(C),panicle number per plant(D),and yield per plant(E)are displayed in histograms.Values are means±SD(n=24).*and**indicate significant differences from the wild type at P ＜0.05 and P ＜0.01,respectively,by Student's t-test.Panicle number is the number of effective tillers that were fertile.Panicle phenotypes(F)include primary and secondary branch numbers and kernel number.Wild indicates the wild type selected from negative lines.

4.Discussion

The crop selection process has created a genetic bottleneck that is restricting breeding output[16,24].Random natural mutations or artificial mutagenesis provide only a limited density of mutational events at uncertain target loci[24].Only a few mutational events which produce numerous genotypes can be selected by natural and artificial selection[25,26].Alternatively,gene editing,such as Cas9,could serve as a localized mutagenic agent and produce high-density mutant populations[15].

In this study,a few novel artificial alleles of the two yieldassociated rice genes Gn1a and DEP1 were generated by the CRISPR/Cas9 system.In particular,one allele of Gn1a and three alleles of DEP1 were superior to those of the natural high-yield alleles.Interestingly,panicle number per plant was the main contributor to the increase in grains per rice plant in the DEP1 mutants(Figs.4-B;S3).This observation differs from previously reported results[3].Given that DEP1 is a pleiotropic gene[3,27],it may cooperate with other genes and be involved in the coregulation of panicle number per plant.However,this speculation needs further verification.

Wild boar: Like the unicorn, the boar appears on several heraldic arms and is closely associated with nobility for it can symbolize ferocity and courage (Biederman 45)

Characterizing and neutralizing similar cases of negative epistasis could improve productivity in many other crops[28].Crop breeding programs targeting yield improvement,resistance to abiotic or biotic stress,and quality have exploited natural superior alleles[16,24].In the task of rice yield increase,novel alleles from this study will make a great contribution.As shown in this study,artificial mutagenesis by gene-editing technology could develop more superior alleles and provide ideal genetic materials for crop breeding. Gene editing increases allelic diversity in functional genes of interest,partly controls biological evolution,and causes the attention of artificial evolution[28,29].The findings from this study reveal the potential of gene editing for easy and rapid improvement of yield-associated traits.

Acknowledgments

This work was supported by the Department of Sciences and Technology of Yunnan Province(2016BB001),the National Basic Research Program of China(2013CB835200)and a Key Grant of Yunnan Provincial Science and Technology Department(2013GA004).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2018.05.005.