Conversion of a normal maize hybrid into a waxy version using in vivo CRISPR/Cas9 targeted mutation activity

Xinto Qi, Ho Wu, Hiyng Jing, Jinjie Zhu, Chngling Hung, Xin Zhng,Chnglin Liu,*, Beijiu Cheng,*

aAnhui Agricultural University,Hefei 230036,Anhui,China

bInstitute of Crop Sciences,Chinese Academy of Agricultural Sciences/National Key Facility for Crop Gene Resources and Genetic Improvement,Beijing 100081,China

ABSTRACT Waxy maize is a specialty maize that produces mainly amylopectin starch with special food or industrial values. The objective of this study was to overcome the limitations of wx mutant allele acquisition and breeding efficiency by conversion of parental lines from normal to waxy maize.The intended mutation activity was achieved by in vivo CRISPR/Cas9 machinery involving desired-target mutation of the Wx locus in the ZC01 background,abbreviated as ZC01-DTMwx. Triple selection was applied to segregants to obtain high genome background recovery with transgene-free wx mutations. The targeted mutation was identified, yielding six types of mutations among progeny crossed with ZC01-DTMwx.The amylopectin contents of the endosperm starch in mutant lines and hybrids averaged 94.9%, while those of the wild-type controls were significantly (P ＜0.01) lower, with an average of 76.9%. Double selection in transgene-free lines was applied using the Bar strip test and Cas9 PCR screening. The genome background recovery ratios of the lines were determined using genome-wide SNP data.That of lines used as male parents was as high as 98.19%and that of lines used as female parents was as high as 86.78%.Conversion hybrids and both parental lines showed agronomic performance similar to that of their wild-type counterparts. This study provides a practical example of the efficient extension of CRISPR/Cas9 targeted mutation to industrial hybrids for transformation of a recalcitrant species.

1.Introduction

Maize(Zea mays L.)is a cereal crop used to produce starch for industrial feedstock and human foods. In a normal maize cultivar, over 70% of the kernel weight is accounted for by endosperm starch [1,2], of which amylopectin accounts for approximately 75% and amylose 25% [3]. Waxy maize, which has a high endosperm amylopectin content,was first discovered in China more than 100 years ago [4]. It was originally consumed by the Asian community and was later consumed worldwide as fresh table ears with ever-increasing demand[5]. Waxy maize also has excellent starch composition for industrial use and high economic value [5,6]. Thus, the acquisition of target gene mutations and increasing waxy maize breeding efficiency should benefit the waxy maize industry.

The Wx locus encodes the enzyme GRANULE BOUND STARCH SYNTHASE I (GBSS I), which is required for amylose synthesis and determines amylose content in both pollen and endosperm tissues [7,8]. In waxy maize, homozygous recessive wx alleles result in maize endosperm composed almost entirely of amylopectin [8,9]. To date, the acquisition of wx mutations are limited to a few natural mutants [3,5].Introgression of recessive mutant alleles using a backcross breeding strategy is hindered by linkage drag effect [10].Generation or acquisition of Wx mutants using more favorable technology such as genome editing should shorten this process.Genome editing using the modified type II bacterium clustered regularly interspaced short palindromic repeats(CRISPR) adaptive immune system can create precise target mutations and has been applied in maize [10-14]. However,heritable mutations obtained by genome editing are highly dependent on the delivery of the machinery to plants via stable transformation [10,14]. Furthermore, stable maize transformation with plant regeneration is highly dependent on the specific genotype and requires laborious and timeconsuming tissue culture[15].Thus,the stable transformation of a few easily transformed genotypes and the use of their genome-editing machinery in trans mode to generate intended mutations remains a practical technique [10] for most important crop species including maize.

The objective of this study was to rapidly transform a normal maize hybrid to a waxy version by mutating the Wx loci of the parental lines using an in vivo desired-target mutator strategy with a previously constructed line having a high-efficiency waxy mutagenesis machinery.

2.Materials and methods

2.1.In vivo desired-target-mutation(DTM)donor and creation of the mutation

The ZC01-DTMwxwas previously developed [9]. The ZC01-DTMwxpollen donor was crossed with 35Fu and 35Mu,the two parent lines of the single-cross hybrid named as 35. The F1,BC1F1, and BC2F1progeny were screened for DTMwxpresence using a bialaphos (Bar) strip test (EnviroLogix, AS-013-LS, ME,US). Approximately 100 BC2F1progeny of each parent line were selfed to obtain the BC2F2generation. All BC2F2plants were triply selected for the wx target mutation, DTMwx-free(transgene-free) status, and genome background recovery.The resulting waxy lines were used to create a waxy version of the single-cross hybrid.

2.2. Detection of genetically modified component

In the F1,BC1F1,and BC2F1generations,positive transformants were identified with the Bar strip test.In the BC2F2generation,transgene-free plants were doubly screened using both the Bar strip test and PCR amplification for negative screening for the Cas9 gene. To increase the rigor of the PCR amplification experiment, both the Bar gene and the Cas9 gene were amplified to identify transgene-free plants. The maize ZmADH1 gene was selected as a reference gene. The primer pair targeting the Cas9 gene was 5′-CACCACCTTCACTGTCTGCA-3′ (forward) and 5′-CAACCGGAAAGTGACCGTGA-3′ (reverse). The primer pair targeting the Bar gene was 5′- GGACACGCTGAAATCACCAG-3′(forward)and 5′-CTTATATGCTCAACACATGAG-3′(reverse).The primer pair targeting the ZmADH1 gene was 5′-CGTCGTTTCCCATCTCTTCCTCC-3′ (forward) and 5′-CCACTCCGAGACCCTCAGTC-3′ (reverse). The PCR cycling parameters were as follows: 95 °C for 10 min followed by 35 cycles of 95°C for 10 s, 60°C for 30 s,and 72°C for 30 s.

2.3. Sanger sequencing to identify the target mutation

Sanger sequencing was used to identify mutations in BC2F2progeny. The genomic regions surrounding ZmWx gene mutation target sites were amplified by PCR using KOD-plus polymerase (KOD-401, TOYOBO Life Science Department,Osaka, Japan) and primer pairs 5′-CATACTTCTCCGGACCATACGGTAA-3′ (forward) and 5′-TCCCTGCTGGGGTCCCACTC-3′ (reverse), spanning the target sites. The PCR amplicons were sequenced with an ABI3730(Applied Biosystems,California,US)to identify the mutations.The ABI sequencing profiles were decoded with DSDecode[16]to identify mutation types.

2.4. Mutant phenotype verification in endosperm starch and pollen grains

Pollen and endosperm starch were stained following Hunt et al. [17]. Mature maize kernels were harvested and soaked in sterile water overnight. The pre-treated endosperm was then scraped onto a glass slide and stained with Lugol solution(10% [w/v] KI, 5% [w/v] I2) diluted 100-fold with water before use. Fresh anthers were treated with Carnoy's reagent(absolute ethyl alcohol:acetic acid 1:1)for 24 h,and the pollen grains were directly stained with Lugol solution. Endosperm starch particles and pollen grains were stained for 5 min and then observed under 10× magnification with an optical microscope (MSHOT,Guangzhou,China).

2.5. Measurement of amylopectin content

Kernel amylopectin content was determined using an amylose/amylopectin assay kit (#K-AMYL, Megazyme, IR) [18].Briefly,endosperm starch samples were completely dispersed by treatment with heated dimethyl sulfoxide and ethanol to remove lipids.The precipitated starch samples were dissolved in an acetate/salt solution.The amylopectin was precipitated with concanavalin A and removed by centrifugation. Finally,the amylose solution and starch acetate/salt solution were enzymatically hydrolyzed to D-glucose, which was determined using the glucose oxidase/peroxidase reagent. GOPOD reagent (Megazyme) was added, and the absorbances of the supernatants of conA-processed samples and total starch samples were measured at 510 nm with a spectrophotometer.

2.6. Whole-genome SNP genotyping and genome background recovery estimation

Genotyping was performed by China Golden Marker Biotech Co.(Beijing,China)using the Affymetrix Microarray CGMB 56K SNP Array [19] containing 56,000 SNPs. The lines, including wildtype ZC01,35Mu,35Fu,and 13 BC2F2mutant lines derived from 35Mu and 35Fu as recurrent parents, were extracted for genomic DNA using the Plant Genomic DNA Kit(DP305,Tiangen Biotech,Beijing,China).The DNA of each line was hybridized in a Hybridization Oven 640 (Affymetrix, State of California, US).Hybridization response was detected using a GeneChip Scanner 3000 (7G) to obtain raw signal CEL files (Affymetrix, California,US). Raw signal CEL files were processed using the Axiom Analysis Suite. All samples passed Dish quality control with a threshold>0.82.SNPs with call rates below 97%were dropped.The variants were classified into seven major categories: Poly High Resolution, No Minor Hom, Hemizygous, Mono High Resolution,Off Target Variant(OTV),Call Rate Below Threshold,and Other. The SNP genotypes same with the recipient line were categorized as the recovered background. A homemade Perl script was used to analyse the genome background recovery rate.A marker distribution map was drawn using the R package R Ideogram[20].

2.7. Field experiment and trait measurements

All lines including the mutant parents, the mutant hybrids,the wild-type parents,and the wild-type hybrids were planted in a randomized block design with three replicates using tworow plots 4 m long and 1.2 m wide. Sixteen plants were retained in each row. To avoid margin effects, five plants in the middle of the row were used to measure plant height,ear height, ear length, kernel row number, kernels per row,hundred-kernel weight, flowering time, and grain yield in each plot. At the maturity stage, plant and ear height were determined.

One-way analysis of variance (ANOVA) and pairwise Student's t-tests of mean values were performed with R 3.3.0. The amylopectin content in the starch of mature seed endosperm of each mutant line was compared to that of the wild type.

3. Results

3.1. The experimental strategy and rationale for the experimental design

To achieve efficient conversion of a normal hybrid into a waxy hybrid, a backcross strategy was applied as illustrated in Fig. 1.The components are as follows. (i) The donor parent was a previously created line [9] containing CRISPR/Cas9 machinery,called DTMwx, that induces in vivo desirable target knockout mutations of ZmWx. (ii) Recurrent parents were either 35Mu or 35Fu,which were respectively the female and male parents of the single-cross hybrid 35. (iii) From the F1to BC2F1generations,DTMwxselection was applied to progeny that showed good growth and grain yield. (iv) Triple selection for the wx target mutation, DTMwxabsence (transgene-free state), and high genome background recovery was imposed in the BC2F2generation. Thus, both parental lines with wx alleles that were transgene-free and had high recovered genetic background were used to make a single-cross hybrid to form the waxy version.

Fig.1- The schematic procedure for efficient creation of DTM on ZmWX line with genetic background recovery.DTMwx indicates lines containing the RNA guided Cas9 DTM machinery,which could induce knockout mutation at ZmWX in vivo.The X and the check sign indicate respectively selected and eliminated plants.

3.2. Generation of desirable ZmWX mutations in two parent recipient lines of a hybrid

Given that the waxy mutant kernel shows a visually distinct fuzzy pattern [4], we selected as candidate plants progeny that showed the mutant kernel phenotype consistently.The selected samples were subjected to target mutation screening. A total of 112 individual plants from eight 35Mu backgrounds and 73 individual plants from five 35Fu backgrounds were selected for Sanger sequencing. Among them, six mutant types were identified(Fig.2).In comparison with the mutant pattern of the mutant donor line ZC01-DTMwx, there were five different mutation types.All six of these mutations were characterized as 1-2 bp short insertions and deletions(InDels),a result consistent with the mutation pattern of the CRISPR/Cas9 machinery[21].

After the triple selection shown in Fig. S1, 13 individual lines were selected (Table 1). Of these, 10 resulted from homozygous mutations of the target gene and three resulted from biallelic knockout.Because the biallelic knockout is also a homozygous recessive mutation irrespective of the segregation of the mutant allele, the trait was wx recessive. The mutant line#35mu3-2(Table 1)carried a 27-bp deletion(Fig.2#6). This deletion would result in the loss of nine amino acid residues in the target region rather than a frameshift knockout mutation. However, subsequent phenotyping still showed a wx recessive mutant phenotype, suggesting that these nine residues are essential for GBSS enzyme activity.

3.3.Identification of transgene-free edited lines and validation of the mutant phenotypes

Removal of the transgenic component of the genome editing machinery is necessary for two reasons.First,the presence of the targeting activity might have some unexpected effect on mutant phenotype identification. Second, having crops that are transgene-free could alleviate public concerns about food safety. To these ends, we applied large-scale screening of mutant candidates using the Bar strip test. Selected mutant lines were verified as transgene-free using both PCR amplification of SpCas9(upper panels of Fig.S1)and the Bar strip test(lower panels of Fig.S1).

Typical examples of mutant pollen grains from the female parent (line #35Mu7-17, Fig. 3-D), the male parent (line#35Fu16-15, Fig. 3-E), and a hybrid line (35Mu7-17 × 35Fu16-15,Fig.3-F)showed markedly weaker staining than their wildtype controls, 35Mu, 35Fu, and 35Mu × 35Fu. Similarly, the endosperm starch of mutants was markedly lower than that of their wild-type controls, which were dark purple-blue (Fig.3-B,G-L).Four sets of female parent,male parent,and singlecross hybrid lines showed higher (P ＜ 0.01) amylopectin contents than those of the wild-type controls. The average amylopectin content of the mutant lines and hybrids was 94.9%, while that of the wild-type controls was 76.9%. Thus,the mutant phenotypes of the created mutants and their single-cross hybrids were verified by multiple methods,including horny transparency, pollen and endosperm KI staining, and determination of the amylopectin content of endosperm starch.

3.4. Genome background recoveries of the mutant lines

Fig.2- Generation of targeted ZmWX mutations using in vivo DTM activity among BC2F2 progeny of two parental lines,35Mu and 35Fu.Nucleotides in red and underlined lettering indicate the protospacer-motif(PAM);Nucleotides in bold blue lettering indicate the spacer region;Nucleotides or dashes in bold red lettering indicate created mutations.

Table 1-Characterization of targeted ZmWX mutations using in vivo DTM activity on progeny of 35Mu and 35Fu, two parents of a single-cross hybrid.

A high proportion of genetic background restoration in biparental lines could preserve the heterotic pattern and thereby the expected agronomic performance of the original hybrid. To evaluate genome recovery, genome-wide highthroughput SNP analysis was applied to the selected mutant lines(Table 2,Fig.4).Based on 35Fu,the male parental line of 35, the mutant line with the highest genetic background recovery, 98.192%, was 35Fu16-15. On average, the five lines created by backcrossing to the male parent showed 95.62%background recovery. In the eight inbred lines created with the female parent of 35Mu, genetic background recovery averaged 88.90%, and the line with highest recovery was 35Mu4-14 with 86.78%. Despite the difference between lines derived from the two parents,all the lines showed recoveries>85%.Genomic heterozygosity rates were between 1.05%and 5.59%(Table 2).Thus,the genetic backgrounds of both parents were largely recovered and the residual heterozygosity was close to that of inbred lines.The parents could be used directly or selfed for a further generation of purification before use as a parent inbred of a hybrid.

3.5. Agronomic performance comparison between the wild type and the created mutants

To determine whether other agronomic traits were changed by the created mutation, we tested the male and female parental lines, their single-cross hybrids, and their corresponding wild types in field trials. The growth traits (Fig. 5,upper panels) plant height, ear height, and grain yield (Fig. 5,lower panels), kernel row number, kernel per row, and hundred-kernel weight, were not significantly different between mutant lines and the wild type (P > 0.05) (Fig. S2). For developmental traits, there was no significant difference in germination, early vigour, flowering, or maturation. Overall,the agronomic performance and plant development among the mutant lines, mutant hybrids, wild-type lines, and wildtype hybrids was similar.

4. Discussion

4.1. Mutant wx alleles in conversion lines were targeted by DTMwx but not mutant allele introgression

In our previous study of ZmLG1, stable expression of CRISPR/Cas9 resulted in a fairly high (> 20%) mutation efficiency [10].The results of the present study support this finding.At least two lines of evidence indicate that the mutant allele in mutant lines was induced by DTMwxand not by mutant allele introgression from ZC01-DTMwxper se. First, there were six mutant wx alleles,of which five were different from the ZC01-DTMwxallele.As shown in Table 1,the same allele as#1(Fig.2,Table 1), the ZC01-DTMwxallele, was found only in two mutant lines, 35Mu7-17 and 35Mu1-7. Only these two lines showed some possibility of mutant introgression from ZC01-DTMwx.However,in the genomic background surrounding the target wx locus neither 35Mu7-17 nor 35Mu1-7 showed genome differentiation from the original mutation donor line,ZC01-DTMwx. We conclude that all 13 mutant lines acquired the target mutations by in vivo DTMwxmutagenesis rather than by mutant-allele introgression.

Fig.3- Characterization of mutant phenotypes for both pollen(A-F)and endosperm starch(G-L, M)among created parental lines and their single-cross hybrids.WT,wild type parental lines and their single-cross hybrid;Edited,the created mutants of the parental lines and their single-cross hybrid.Pollen grains(A-F,Bar=200 μm)and endosperm starch(G-L, Bar=50 μm)potassium iodide(10%KI)staining profiles;(A,G)35Mu(wild type);(B,H)35Fu(wild type);(C,I)35Mu×35Fu;(D,J)35Mu7-17;(E,K)35Fu16-15;(F,L)35Mu7-17×35Fu16-15.M,Amylopectin content(%)comparison between four pairs of created mutant parental lines,their hybrids,and their original wild type lines and hybrids(wild type).a and b, statistical groups by t-test at P ＜0.01.

In addition, we have found a few non-frameshifted alleles such as #6 in Fig. 2. However, the same phenotypes with frameshifted alleles were observed, as shown by 35Mu3-2 in Table 1. These lines with non-frameshifted alleles offer potential for further elucidating the structure of the enzyme GBSS I.

4.2. Creation of mutant parental lines with different genome recovery proportions

As proposed in our previous publication but without demonstration[10],we have further verified in the present study the rapid background recovery and the overcoming of linkage drag afforded by the in vivo DTM strategy. Whole genome recovery estimates revealed a difference in the average genetic background recovery of the two parental lines. The male parent recovery rate slightly exceeded 95%; however, the female parental recovery rate was approximately 87%. The reason for this difference was probably due to the closer relationship of the male than the female parent with the donor line, ZC01-DTMwx. The genetic similarity between 35Fu and ZC01-DTMwxreached 66.1%, while the genetic similarity between 35Mu and ZC01-DTMwxwas 60.5%. Although the rate of the female parent line was lower, the overall rate reached 87%. This finding is in close accord with the 87.5% genetic model of the BC2F2breeding scheme. Thus, the results obtained from whole-genome SNP analysis are reasonable and credible. The agronomic traits, including plantarchitecture, appearance, height, growth stages, and grain yield, also support the findings of the genome recoveries of the mutant lines and their hybrids(Figs.4,5, Table 2).

Table 2-Genome recovery evaluated by genome-wide SNP analysis.

As mentioned above, wx genotyping indicated that the mutated genes resulted from in vivo mutagenesis mediated by ZC01-DTMwx rather than from gene introgression. SNP genotyping at both ends of the wx locus also supported this result. The neighboring regions of the wx gene(chr9:23,107,329-23,111,318) represented by the SNPs #AX-86254404 (chr9:23,095,142) and #AX-86320759(chr9:23,397,073) were identical to those of the recipient genome (Table S1, Fig. 2). In a high-resolution recombination map [22], the crossover events per chromosome were as few as two during maize meiosis. As a result of rare crossovers during meiosis,large donor DNA segments from donor should be retained as a result of linkage drag when a conventional backcrossing program is adopted.This,the backcross strategy employing in vivo targeted mutation activity using ZC01-DTMwxyielded an advantage in genome recovery efficiency over conventional backcross breeding.

Fig. 4 - Genome background recoveries of an example pair of bi-parental lines.

Fig.5- Comparison of agronomic performance between wild type and created mutant bi-parental lines and single-cross hybrids.Upper panels,pollinated plants,Bar,20 cm;lower panels,mature ears.Bar,5 cm.

4.3.Agronomic performance of the mutants and the wild type

Another reason for the similar agronomic performance is that the target gene, Wx, is one of the genes involved in starch metabolism downstream of the metabolic pathway.Thus,this gene has no regulatory effect on other genes and very few side effects, so that the phenotype is very similar for the mutant and the wild type, except for the target trait of the gene. In contrast,our other example[10]of mutation of ZmLG1,an SPL family transcription factor, resulted in a large difference in agronomic performance because thousands of genes in downstream regulatory networks were changed by this induced mutation(results not shown).

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2020.01.006.

Declaration of competing interest

The authors declare no conflicts of interest.

Acknowledgments

This work was supported the National Transgenic Science and Technology Program(2019ZX08010-003),the National Natural Science Foundation of China(31771808),the National Key Research and Development Program of China (2016YFD0101803), the Key Area Research and Development Program of Guangdong Province(2018B020202008), Beijing Municipal Science and Technology Commission(D171100007717001),and National Engineering Laboratory for Crop Molecular Breeding.

Author contributions

B.Cheng and C.Liu conceived the study.X.Qi,H.Wu,H.Jiang,J.Zhu,C.Huang,and X.Zhang designed the experiments.X.Qi and H. Wu performed the experiments. H. Jiang, J. Zhu, C.Huang, and X. Zhang supervised the study. B. Cheng, C. Liu,and X.Qi wrote the manuscript.