*2.5. Expression Analysis of Cas9 in Transgenic Plants*

The expression level of *Cas9* was examined in transgene-positive plants of several randomly selected T1 lines by qRT-PCR. The *Cas9* mRNA level was approximately 11.6–30.3 fold that of the *OsActin1* gene in different lines (Figure 6).

**Figure 6.** qRT-PCR analysis of *Cas9* expression in the transgene-positive plants of the T1 lines. The rice *Actin1* gene was used as the internal control. Error bars represent standard deviations of three biological repeats.

#### **3. Discussion**

In this study, rice *FWL* family genes were mutated using the *Agrobacterium*-mediated CRISPR/Cas9 system, and the phenotypes of mutants of two genes (*OsFWL1* and *OsFWL4*) were characterized. The results suggest that the *OsFWL4* gene is a negative regulator of tiller number and plant yield in rice and that the *OsFWL1* gene plays a role in modulating rice grain length.

Rice tiller number is an important agronomic trait that largely affects grain yield. The tiller number of the *OsFWL4* gene mutants was increased by up to 45.9% compared with that of the WT (Figure 2A,B). Additionally, flag leaf width of mutants was also increased (Figure 3A,C). Leaf epidermal cell observation revealed that the increase in leaf width of mutants was caused by an increase in cell number but not in cell size (Figure S4). Hence, OsFWL4 may negatively affect cell proliferation during leaf and tiller development. In the mutants, the grain yield per plant was increased by up to 35.8% (Figure 2E), suggesting that the *OsFWL4* gene may be useful in breeding to improve rice yield.

The grain length of the *OsFWL1* gene mutants was significantly higher than that of the WT (Figure 5A,B). Similarly, grain length of the *OsFWL3* gene mutant has also been reported to be increased [9]. However, grain width of the *OsFWL1* gene mutants was not affected and grain thickness was reduced (Figure 5C,D). The decrease in grain thickness might be caused by insufficient grain filling due to enlarged glumes in the mutants. Finally, grain weight of mutants was not changed. Together, these results suggest that rice *FWL* genes play a role in the regulation of organ development in rice.

It has been reported that the *OsFWL4* gene can enhance Cd resistance when expressed in yeast cells and mediate the translocation of Cd from the roots to shoots in rice seedlings [15]. Recently, the *OsFWL1* gene was also reported to mediate Cd homeostasis in rice [20]. Hence, the two rice

*FWL* genes function in both organ development regulation and Cd homeostasis in rice. Similarly, the *OsFWL5*/*PCR1* gene has been found to confer Cd resistance and Zn hypersensitivity upon expression in yeast and to modulate grain size and weight and metal ion homeostasis in rice [13,14]. However, how a single *FWL* gene fulfills such diverse roles remains unknown. Interestingly, the OsFWL5/PCR1 protein has been found to be localized as oligomers in the plasma membrane microdomains [13]. Additionally, GmFWL1, an important FWL protein involved in soybean nodulation, has also been demonstrated to be a plasma membrane microdomain-associated protein [5,10]. The plasma membrane microdomains are membrane sub-compartments consisting of special lipids and proteins and are considered signal integration hubs of cells [35]. Hence, the membrane microdomain-associated FWL protein may act in several distinct signaling pathways and thus affect multiple biological processes in plants. Both OsFWL1 and OsFWL4 proteins are located in the plasma membrane, and the OsFWL4 protein is distributed in a punctate manner [9,15]. We speculate that OsFWL4 may also be a microdomain-associated protein. A gene co-expression analysis revealed that *OsFWL4* may be involved in many cell functions (Tables S7 and S8). The *OsFWL1* gene is reported to be co-expressed with the zinc finger and ubiquitination-related protein genes [9].

CRISPR/Cas9 DNA can be delivered into rice cells by *Agrobacterium*-mediated transformation and integrated into the rice genome. Studies have shown that T-DNA truncation frequently occurs in *Agrobacterium*-mediated transformation [36,37]. Detection of CRISPR/Cas9 DNA in unmutated T0 plants revealed that most (25 out of 39, excluding two plants that escaped hygromycin selection) of them lacked sgRNA and/or *Cas9* transgenes (Table S2). This indicates that the integrity of the sgRNA/*Cas9* expression cassette is an important factor affecting editing efficiency. Truncation of T-DNA can occur at its different ends (left, right, or both ends) and different stages of integration (before or during integration) [36–39]. In rice, truncated T-DNAs were detected in more than 18% of the transformants [40]. Hence, improving the quality of T-DNA integration may aid in further increasing the efficiency of CRISPR/Cas9 gene editing based on *Agrobacterium*-mediated transformation.

T-DNA segregation analysis revealed that transgene-free plants could be obtained in several T1 plants for all lines examined (Figure S2). This suggests that T-DNA insertions in CRISPR/Cas9 gene-edited plants can be easily eliminated in the T1 generation. Interestingly, inconsistent segregation of *HPT*, sgRNA, and *Cas9* transgenes was observed in two lines (Osfwl3a#4 and Osfwl4b#6; Figure S2). The absence of sgRNA and *Cas9* transgenes in Osfwl4b#6 T1 plants was caused by the lack of these sequences in the T0 plant. In the 20 T1 plants examined for line Osfwl3a#4, seven plants contained only the sgRNA transgene, 12 plants contained all the three transgenes, and one plant had no transgene (Figures S2 and S3). This inconsistent segregation could be attributed to the presence of two T-DNA insertion sites in this line; one contained the complete T-DNA fragment, whereas the other harbored a truncated T-DNA with only the sgRNA transgene.

The off-target effect is a major concern in the application of CRISPR/Cas9 technology. Several studies have reported that the CRISPR/Cas9 system is highly specific in plants [27,32,41,42]. However, moderate or even high-frequency off-target mutagenesis has also been reported [43–46]. In the present study, potential editing events at 30 putative off-targets of the 15 sgRNAs were examined. We detected mutations in 13 out of the 30 putative off-target sites (Table 3). Analysis of the relationship between mismatch numbers of target-like sequences and off-target activity revealed that all the sequences harboring single mismatches with the sgRNAs and 80.0% of the sequences containing double mismatches were cleaved. These results indicate that at least two mismatches between the sgRNA and potential off-target sequences are required to minimize the off-target effects. Interestingly, an off-target site with up to four mismatches (Osfwl2aOFF-2) was also mutated (Table 3). The first mismatch of this site located at the first base in the 5 end is usually tolerated by CRISPR/Cas9. Additionally, all four mismatched bases of this site were adenine (Table 3), which led to rN:dT base pairing during sgRNA binding. Generally, the rN:dT mismatches are well tolerated [47,48]. Hence, both the identity and position of mismatched bases might contribute to the cleavage of this site by Cas9. The results suggest that the sgRNAs should be designed carefully to minimize or avoid off-target mutagenesis in plants.
