Control of Grain Shape and Size in Rice by Two Functional Alleles of OsPUB3 in Varied Genetic Background

Grain shape and size are key determinants of grain appearance quality and yield in rice. In our previous study, a grain shape QTL, qGS1-35.2, was fine-mapped using near-isogenic lines (NILs) derived from a cross between Zhenshan 97 (ZS97) and Milyang 46 (MY46). One annotated gene, OsPUB3, was found to be the most likely candidate gene. Here, knockout and overexpression experiments were performed to investigate the effects of OsPUB3 on grain shape and size. Four traits were tested, including grain length, grain width, grain weight, and the ratio of grain length to width. Knockout of OsPUB3 in NILZS97, NILMY46, and another rice cultivar carrying the OsPUB3MY46 allele all caused decreases in grain width and weight and increases in the ratio of grain length to width. Results also showed that the magnitude of the mutational effects varied depending on the target allele and the genetic background. Moreover, it was found that NILZS97 and NILMY46 carried different functional alleles of OsPUB3, causing differences in grain shape rather than grain weight. In the overexpression experiment, significant differences between transgenic-positive and transgenic-negative plants were detected in all four traits. These results indicate that OsPUB3 regulates grain shape and size through a complex mechanism and is a good target for deciphering the regulatory network of grain shape. This gene could be used to improve grain appearance quality through molecular breeding as well.


Introduction
Rice (Oryza sativa L.) provides a staple food source for more than half the world's population. Grain shape and size are important appearance quality of rice, which are one of the most direct characteristics for consumers and influence the market value of grain products. In general, long and slender grains have higher competitive market value, which is preferred by consumers in most regions of the world [1][2][3][4]. Grain size is also a major determinant of grain weight, one of the three yield components (number of panicles per plant, number of grains per panicle, and grain weight). Therefore, understanding the genetic basis of grain shape and size is vital for improving grain quality and yield of rice.
Grain shape and size are largely determined by grain length and width, which are controlled by a large number of quantitative trait loci (QTL). At least 25 QTLs for grain shape and size in rice have been cloned. Fifteen of them mainly regulated grain length, including qTGW1.2b [5], GS2/GL2 [6,7], OsLG3 [8], OsLG3b/qLGY3 [9,10], GS3.1 [11], GS3 [12], SG3 [13], GL3.1/qGL3 [14,15], qTGW3 [16], qGL5 [17], TGW6 [18], GW6a [19], GL6 [20], GLW7 [21] and GL10/OsMADS56 [22,23]. Seven other QTLs mainly regulated grain width, including GW2 [24], TGW2 [25], GS5 [26], GSE5 [27], GW6 [28], GW8 [29] and GW10 [30]. The remaining three genes, including GSA1 [31], GL7/GW7 [32,33], and The CRISPR/Cas9 system was used to produce mutants for validating the effects of OsPUB3. OsPUB3 protein contained a U-box motif and Armadillo (ARM) repeats in the central region. A site located in the distal N-terminal region was selected as the target for CRISPR/Cas9 gene editing ( Figure 1A). To compare the effects of the two parental alleles, OsPUB3 ZS97 and OsPUB3 MY46 , and to test the effects of OsPUB3 in different genetic backgrounds, three recipients were used. They are NIL ZS97 and NIL MY46 previously used for fine-mapping qGS1- 35.2, and another indica rice cultivar carrying the OsPUB3 MY46 allele, Zhonghui 161 (ZH161). NIL ZS97 and NIL MY46 have a small difference in grain shape and size, whereas the grain of ZH161 is much thinner and smaller ( Figure 2).   OsPUB3. OsPUB3 protein contained a U-box motif and Armadillo (ARM) repeats in the central region. A site located in the distal N-terminal region was selected as the target for CRISPR/Cas9 gene editing ( Figure 1A). To compare the effects of the two parental alleles, OsPUB3 ZS97 and OsPUB3 MY46 , and to test the effects of OsPUB3 in different genetic backgrounds, three recipients were used. They are NIL ZS97 and NIL MY46 previously used for fine-mapping qGS1-35.2, and another indica rice cultivar carrying the OsPUB3 MY46 allele, Zhonghui 161 (ZH161). NIL ZS97 and NIL MY46 have a small difference in grain shape and size, whereas the grain of ZH161 is much thinner and smaller ( Figure 2).   A total of twelve independent T 0 mutants were obtained, including three homozygous and nine biallelic mutants ( Figure 1B). For NIL ZS97 , two homozygous mutants and one transgenic-negative control were identified. KO-ZS-1 had a 1 bp deletion and KO-ZS-2 had a 1 bp insertion. For NIL MY46 , one homozygous and six biallelic mutants were identified. No negative transformant was found. The homozygous mutant had a 1 bp insertion, and each of the biallelic mutants contained a 1 bp insertion and a 1 bp deletion. For ZH161, three biallelic mutants and one negative transformant were identified. KO-ZH-1 contained a 1 bp insertion and a 1 bp deletion, KO-ZH-2 contained a 1 bp insertion and a 22 bp deletion, and KO-ZH-3 had 22 bp and 35 bp deletions. The mutations all led to a frameshift and produced premature stops before the U-box motif and ARM repeats ( Figure S1).

Phenotypic Change Due to OsPUB3 Knockout
Phenotypic changes resulting from the knockout of OsPUB3 were tested in 2021. Three sets of rice materials were used: (a) recipient NIL ZS97 , T 1 lines of the two homozygous mutants, and a transgenic-negative control (CK ZS97 ); (b) recipient NIL MY46 , a T 1 line of the homozygous mutant, and six T 1 populations segregating the biallelic mutations; (c) recipient ZH161, six T 2 homozygous mutants derived from the three biallelic mutants, and a transgenic-negative control (CK ZH161 ). All these populations were measured for four traits of grain shape and size, including grain length (GL), grain width (GW), 1000 grain weight (TGW), and the ratio of grain length to width (RLW). The ZH161-type lines were additionally measured for other five traits, including the number of panicles per plant (NP), number of spikelets per panicle (NSP), number of grains per panicle (NGP), grain yield per plant (GY), and heading date (HD). Student's t-test was performed to analyze phenotypic differences between mutants and controls.
The two homozygous mutants of NIL ZS97 both showed significant decreases in GL, GW, and TGW compared with the recipient and CK (Table 1). Decreases in KO-ZS-1 and KO-ZS-2 over CK ZS97 were 0.102 and 0.204 mm for GL, 0.080 and 0.122 mm for GW, and 2.14 and 2.67 g for TGW, respectively. The CK itself decreased over the recipient NIL ZS97 on these traits. Thus, decreases in the two mutants over NIL ZS97 were larger, becoming 0.266 and 0.369 mm for GL, 0.091 and 0.133 mm for GW, and 3.02 and 3.55 g for TGW.
Because changes in GL and GW had the same direction, their influences on the GL/GW ratio were mitigated. The two mutants showed non-significant differences with NIL ZS97 and significant increases over CK ZS97 .
The NIL MY46 mutants were only tested against the recipient due to unavailable transgenic CK. In each of the six segregating populations, non-significant phenotypic variation was found among the three genotypic groups (data not shown); thus, data of the three genotypes were merged. Together with the homozygous mutant KO-MY-1, a total of seven mutants were analyzed (Table 1). Compared with NIL MY46 , the mutants all showed decreases in GL, GW, and TGW, which were significant except for GL in KO-MY-1 and KO-MY-3. The decreases in GL, GW, and TGW ranged from 0.010-0.270 mm, 0.065-0.218 mm, and 1.72-4.23 g, and averaged 0.124 mm, 0.143 mm, and 3.02 g, respectively. Compared with the differences between NIL ZS97 and its mutants, these effects had the same direction, and the decreases were smaller on GL, larger on GW, and similar on TGW. Accordingly, increases in the GL/GW ratio over the control were much larger in the mutants of NIL MY46 than that of NIL ZS97 .
Among the six homozygous mutants of ZH161, all and five showed significant decreases in GW and TGW over the two controls. Compared with ZH161 and CK ZH161 , the decreases were averaged as 0.093 and 0.068 mm for GW and 0.052 and 0.072 g for TGW, respectively (Table 1). For GL and RLW, four and six mutants showed significant increases over the two controls. Compared with ZH161 and CK ZH161 , the increases were averaged as 0.215 and 0.114 mm for GL and 0.184 and 0.111 for RLW, respectively. The mutational directions, compared with that detected in the NIL ZS97 and NIL MY46 populations, were the opposite for GL but the same for the other three traits. Variations among the three experiments were also found in the magnitude of mutational effects on GW, TGW, and RLW. In accordance with the thinner and smaller grains of ZH161 than NIL ZS97 and NIL MY46 , the knockout caused fewer decreases in GW and TGW when using ZH161 as the recipient, especially in TGW. For RLW, knockout on the OsPUB3 MY46 allele carried by ZH161 and NIL MY46 caused much larger increases than on the ZS97 allele carried by NIL ZS97 . As for the five traits that were only tested in the ZH161 population, the most consistent results were found in the differences in NP, NSP, and NGP between mutants and CK ZH161 . General decreases in NP and increases in NSP and NGP were observed (Table S1).  In summary, the knockout of OsPUB3 in the three recipients all resulted in increased RLW and decreased GW and TGW, whereas the effects on GL were less consistent. Meanwhile, the magnitude of the mutational effects on GW, TGW, and RLW varied depending on the target allele as well as the genetic background. NIL MY46 is the only recipient on which OsPUB3 knockout caused large changes for all the RLW, GW, and TGW.

Effects of Expressing OsPUB3 with Rice Actin 1 Promoter
Overexpression transgenic plants were generated to further validate the effects of OsPUB3. Since the knockout of OsPUB3 in NIL MY46 caused the largest effect, we amplified the coding sequence of OsPUB3 from MY46 and introduced it into NIL MY46 driven by a rice Actin 1 promoter. Four independent T 1 populations, namely OE-1 to OE-4, were selected to measure the four grain-shape traits. The Student's t-test was performed to determine phenotypic differences between negative and positive transgenic plants in each T 1 population.
Significant differences were detected for all four traits in at least two populations ( Table 2). For GL, a significant increase in positive plants over negative plants was observed in each population. The increases ranged from 0.049 to 0.104 mm. For GW, a decrease in positive plants over negative plants was shown in each population but only significant in two populations, OE-1 and OE-4, reduced by 0.039 and 0.042 mm, respectively. For TGW, a significant increase in positive plants over negative plants was detected in two populations, OE-2 and OE-4, rising by 0.84 and 0.46 g, respectively. The other two populations showed a non-significant increase or decrease. For RLW, an increase in positive plants over negative plants was observed in each population, but only significant in three populations, OE-1, OE-3, and OE-4. The significant increases ranged from 0.051 to 0.074. These results showed that overexpression of the OsPUB3 MY46 allele in NIL MY46 increased GL but decreased GW. Due to the effect being larger on GL than on GW, increases were observed for RLW and TGW.

Discussion
Most important agronomic traits are controlled by a few major QTLs and many minor QTLs. Only a small number of genes underlying minor QTLs have been isolated in rice [5,13,23,48,49]. In our previous study, minor QTL qGS1-35.2 was fine-mapped using NIL populations derived from a cross between ZS97 and MY46. One annotated gene, OsPUB3, was found to be the most likely candidate gene. In these NIL populations, the ZS97 allele had a slight effect causing larger GL and smaller GW compared to the MY46 allele, thereby qGS1-35.2 exhibited a significant effect on RLW [47]. In the present study, the effect of OsPUB3 was investigated through knockout and overexpression experiments. Knockout of OsPUB3 in NIL ZS97 , NIL MY46 , and ZH161, another cultivar carrying the OsPUB3 MY46 allele, caused consistent effects on RLW, GW, and TGW, as well as significant but less consistent effects on GL. These effects, reflecting differences between function and lossof-function alleles, are much larger than those detected in the fine-mapping of qGS1-35.2 that reflected the differences between the functional OsPUB3 ZS97 and OsPUB3 MY46 alleles.
Comparing the mutational effects on the four traits between NIL ZS97 and NIL MY46 , changes of 3.28 and 3.02 g in TGW are the closest, and that of 0.000 and 0.098 in RLW are the furthest. This is in agreement with the previous result that the effect of qGS1-35.2 was significant on grain shape and non-significant on grain weight. These results also confirm our previous hypothesis that the small effect of a minor QTL is a result of the small contrast between partially function alleles carried by the parental lines [5]. There is abundant genetic variation in natural populations, and different alleles of a gene may have different functions [50][51][52]. The identification and screening of different alleles are helpful to rice improvement.
Overexpression and knockout/knockdown of a same gene usually causes opposite effects [10,11,13,15,17,20,23,25,27,28,31,34]. However, we surprisingly discovered that overexpressing and knockout OsPUB3 MY46 both decreased GW. This implied that OsPUB3 regulated GW by a rare molecular mechanism. OsPUB3 encodes a U-box type E3 ubiquitin ligase, which has been verified to have ubiquitination activities [53]. There are very few examples of overexpression and knockout/knockdown of the same gene exhibiting a similar phenotype [54][55][56][57][58][59][60][61]. Interestingly, three of them were found to be involved in the ubiquitin-proteasome pathway. SPL35 regulated rice defense response by interacting with a ubiquitin-conjugating E2 enzyme. Both overexpression and knockdown of SPL35 caused the lesion to mimic the phenotype [61]. UNC-45 functioned the organization of myosin. Loss of function of UNC-45 resulted in paralyzed animals. The overexpressing UNC-45 promoted the nonnative myosin conformation that was degraded by ubiquitination complexes, displaying a paralysis phenotype as well [55,57,59]. Usp28 encoded a deubiquitinase that stabilized the subunit of SCF ubiquitin ligase, Fbw7. At the highest level, Usp28 could also stabilize substrates of Fbw7, proto-oncogenes. Knockout of Usp28 triggered Fbw7 degradation and accumulation of Fbw7 substrates, resulting in oncogenic transformation. Overexpressing Usp28 stabilized both Fbw7 and its substrates, also causing oncogenic transformation [58]. These examples illustrated that ubiquitination exquisitely regulates protein turnover and homeostasis through a complex interaction system, though the regulatory mechanisms of these genes were different. Whether OsPUB3 regulated GW through one of the similar molecular mechanisms described above or through other mechanisms remains to be explored.
Our results also showed that the effects of OsPUB3 varied depending on genetic background. Knockout of OsPUB3 MY46 decreased GL in the NIL background but increased the trait in the ZH161 background. As for GW, the magnitude of the mutational effect was much larger in the NIL background than that in the ZH161 background, though knockout OsPUB3 MY46 always exhibited decreasing effects. Consequently, the decrease in TGW was approximately five-fold greater in the NIL background than that in the ZH161 background, while the increase in RLW in the NIL background was only half of that in the ZH161 background. The grains of ZH161 were much thinner and smaller compared to those of NIL. The effect variation of OsPUB3 in the two backgrounds may be related to the QTLs responsible for the difference in grain shape between the two varieties. The genetic interaction between OsPUB3 and these QTLs would be a good starting point to investigate the molecular mechanisms underlying the role of OsPUB3 in regulating grain shape and size. Overall, OsPUB3 regulates grain development through a complex mechanism and is a good target gene for deciphering the regulatory network of grain shape and size in rice.
Grain shape is an important appearance quality, which greatly influences the market value of rice products. Large RLW generally confers better appearance quality and competitive market value. Some breeding programs recently even attempted to convert the traditional short and bold grains of japonica rice into long and slender grains because of their excellent appearance quality [55,56]. Due to a more obvious reduction in grain width or increase in grain length, overexpressing OsPUB3 and knockout in different genetic backgrounds always increased the RLW. Moreover, OsPUB3 seems to coordinate the trade-off between grain weight and other yield traits. The knockout of OsPUB3 reduced grain weight but increased the number of panicles or grains, resulting in a stable grain yield. Therefore, OsPUB3 could be used to improve grain appearance quality without yield penalty through genetically modified breeding.

Knockout OsPUB3 in Three Rice Cultivars
The CRISPR/Cas9 system was used to knockout OsPUB3. One target, located at +74 to +93 in the coding region ( Figure 1A), was selected using the web-based tool CRISPR-GE (http://skl.scau.edu.cn, accessed on 6 March 2018). The oligonucleotides 3900-cri (Table S2) were designed and ligated into the BGK03 vector (BIOGLE Co., Ltd., Hangzhou, China) according to the manufacturer's instructions. The original BGK03 vector contains a rice U6 promoter for activating the target sequence, a Cas9 gene driven by the maize ubiquitin promoter, and a hygromycin marker gene driven by the Cauliflower mosaic virus 35S promoter.
Three transgenic recipients were used, including NIL ZS97 , NIL MY46 , and ZH161. The NIL ZS97 and NIL MY46 were derived from the cross between ZS97 and MY46, which were previously used for fine-mapping qGS1-35.2. ZH161 is an indica rice cultivar carrying the OsPUB3 MY46 allele. The ORF sequence of OsPUB3 in ZH161 was analyzed according to the following procedure. Genomic DNA was extracted from a 2 cm long leaf sample of ZH161 using DNeasy Plant Mini Kit (QIAGEN, Hilden, German). Full-length ORF of OsPUB3 was amplified using the primer pairs 3900-seq (Table S2) and the product was sequenced by the Sanger method.
The CRISPR/Cas9 constructs were separately introduced into three recipients using Agrobacterium tumefaciens-mediated transformation. Genomic DNA of T 0 plants was extracted from a 2 cm long leaf sample using the DNeasy Plant Mini Kit. They were assayed with the hygromycin gene marker Hyg (Table S2). The OsPUB3 gene fragment was amplified from each Hyg-positive plant using the primer pairs 3900-cri-seq (Table S2). The product was directly sequenced by the Sanger method and decoded using the webbased tool DSDecodeM (http://skl.scau.edu.cn/dsdecode, accessed on 10 May 2019 and 27 July 2020).

Expressing OsPUB3 with Rice Actin 1 Promoter
Total RNA was extracted from leaves of MY46 using RNeasy Plus Mini Kit (QIAGEN, Hilden, German). The 1 st strand cDNA was synthesized using ReverTra AceR Kit (Toyobo, Osaka, Japan). The OsPUB3 cDNA fragment that contained the full coding sequence was amplified using the primer pairs 3900-oe (Table S2). The product was recombined in the pCAMBIA2300 vector using an In-Fusion Advantage Cloning kit (Clontech, Osaka, Japan). The original pCAMBIA2300 vector comprised a rice Actin1 promoter for activating the target sequence and a neomycin marker gene driven by CaMV 35S promoter. The overexpression construct was introduced into NIL MY46 using Agrobacterium tumefaciensmediated transformation. Total DNA was extracted from a 2 cm long leaf sample of transgenic plants using the method of Zheng et al. [62]. Genotypes of transgenic plants were assayed with the neomycin gene marker Neo marker (Table S2).

Field Experiments and Phenotyping
All the rice materials were tested in 2021 at paddy field in the China National Rice Research Institute in Hangzhou, Zhejiang Province, China. Seeds were sown in paddy fields in May and raised in wet bed condition for 26 days. Then, the plants were transplanted with spacing of 16.7 cm between plants and 26.7 cm between rows. In each population, the plants were randomly planted. Normal agricultural practice was employed in field management. From transplanting to harvesting, irrigation was applied to maintain a wellwatered condition except that the water was drained for five days at the maximum tillering stage and before harvesting.
For knockout experiment in NIL ZS97 , each of the four T 1 homozygous lines contained 10 plants (Table 1). For knockout experiment in NIL MY46 , 12 plants were planted for the control NIL MY46 and 36 T 1 plants were grown for each of the six segregating populations and one homozygous line. Genotyping was performed using the marker 3900-cri-seq, and those showing an ambiguous genotype were removed. For knockout experiment in ZH161, each of the eight T 2 homozygous lines contained 30 plants. For overexpression experiment, 30 T 1 plants were grown for each population. Genotyping was performed using the Neo marker.
In October, plants in each population were individually harvested at maturity for phenotyping. Four traits of grain shape and size, including GL, GW, TGW, and RLW, were measured in all experiments. Approximately 6 g of fully filled grains were divided into two halves and measured for the four traits using an automatic seed counting and analyzing instrument (Model SC-G, Wanshen Ltd., Hangzhou, China). Other five traits, including NP, NSP, NGP, GY, and HD, were additionally measured in the ZH161 knockout experiment. The Student's t-test was performed to determine phenotypic differences between mutants and controls in the knockout experiments and between negative and positive transgenic plants in the overexpression experiment.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/plants11192530/s1, Figure S1: Amino acid sequences of OsPUB3 in the three recipients (NIL ZS97 , NIL MY46 , and ZH161) and their homozygous mutants; Table S1: Four yield traits and heading date in ZH161, the transgenic-negative control, and knockout mutants of OsPUB3; Table S2: Primers used in this study.