Splicing of precursor mRNAs (pre‐mRNAs) is a critical biological process of gene expression. Normally, pre‐mRNAs are processed by spliceosomes to produce mature mRNAs by removing introns at 5′‐(donor) and 3′‐(acceptor) splice sites based on the canonical GU‐AG rule (Reddy et al., 2013). Mutation at either the intron donor or acceptor sites should cause mRNA mis‐splicing. Recently, an exon skipping method has been developed to generate loss of gene function in mammalian cells using base editors to mutate nucleotides at the acceptor sites (Gapinske et al., 2018). Moreover, it has been approved that CRISPR/Cas9‐mediated exon skipping in rabbits depends on non‐sense‐associated altered splicing induced by premature termination codon (PTC) mutation (Sui et al., 2018). In contrast, triggering exon skipping by genome editing in plants is still controversial as disruption of the intron donor or acceptor sites in plant genes did not trigger exon skipping but allowed either intron retention or aberrant splicing (Li et al., 2019), following the fact that intron retention is the most frequent mode of alternative splicing in plants (Ner‐gaon et al., 2004).

Fragrance is one of the most important rice quality traits (Abdelrahman and Zhao, 2020). The natural aroma substance 2‐acetyl‐1‐pyrroline (2AP) is the major contributor to the aroma flavour of fragrant rice. The rice gene OsBADH2 encodes a betaine aldehyde dehydrogenase (BADH) that inhibits 2AP biosynthesis. Mutations in OsBADH2 could result in 2AP accumulation and rice fragrance (Chen et al., 2008). To endow the indica rice cultivar R317 with fragrance, we adopted CRISPR/Cas9 to mutate the OsBADH2 gene. A sgRNA targeting immediate upstream of the splice donor site (GT) of the intron 2 was designed (Figure 1a, b) and the vector pYLCRISPR‐Cas9Pubi‐H_BADH2 was introduced into R317 by Agrobacterium‐mediated transformation. Two types of mutants (M190‐5 and M190‐13) were retrieved in the T0 generation plants (Figure 1c). M190‐5 is a homozygous mutant with a single‐nucleotide deletion (ΔG) at the exon–intron junction immediately upstream of the 5′‐splice site of intron 2, while M190‐13 is a bi‐allele edited mutant with a single‐nucleotide deletion (ΔG) and a trinucleotide deletion (ΔAAG).

The homozygous mutation in M190‐5 was faithfully transmitted to T1 generation, while the heterozygous mutations in M190‐13 were transmitted to the T1 generation following the Mendelian law of segregation, resulting in three genotypes: ΔG/ΔG, ΔG/ΔAAG and ΔAAG/ΔAAG. Homozygous T1 plants were analysed for the presence of Cas9 and/or off‐target mutations. From 88 T1 plants tested, 6 plants were free of Cas9 gene and with no off‐target mutation, belonging to two genotypes (Rbadh2^ΔG and Rbadh2^ΔAAG; Figure 1c).

The OsBADH2 gene encodes a functional protein with 503 amino acids (Chen et al., 2008). To check the alteration of the OsBADH2 mRNA in the edited plants, RT‐PCR with a forward and a reverse primer in exons 1 and 6, respectively, was conducted to amplify the cDNAs from the homozygous mutants (Rbadh2^ΔG and Rbadh2^ΔAAG) and the wild type (WT). A 528bp‐cDNA fragment was amplified from the mutant plants, while the WT generated a 671bp‐fragment as expected (Figure 1d). Sanger sequencing revealed that the cDNA fragment from the WT is the expected size of exons of OsBADH2, indicating that the spliceosome recognizes the 5′‐splice sites and cleaves the sites following normal splicing process. To our surprise, not only the introns but also the exon 2 were absent in the 528bp‐cDNA from the edited plants (Figure 1e, f), indicating that the deletion of nucleotides (ΔG or ΔAAG) caused an exon skipping splicing pattern (Figure 1f). Apparently, the alteration (ΔG or ΔAAG) at the donor site of intron 2 prevented the spliceosome from normal splice site recognition. The Osbadh2 mRNA missing exon 2 from Rbadh2^ΔG and Rbadh2^ΔAAG proves that the exonic nucleotide (G) immediately upstream of the exon–intron junction is critical for the normal splicing process. Deletion of this exonic nucleotide caused the entire exon skipping. Recently, it has been shown that base editing‐mediated disruption of either the native intron donor site or acceptor site in plant genes caused intron retention or mis‐spliced segments, but no exon skipping mutation (Li et al., 2019). Our present study showed that exon skipping could happen to occur in plants by CRISPR/Cas9‐mediated mutations at the exon end immediately upstream of the exon–intron junction. However, it is unknown that whether the deletion of exonic nucleotide upstream of the intron donor site can always lead to exon skipping in plants.

It has been reported that base editing‐mediated exon skipping depends on PTC mutations in rabbits (Sui et al., 2018). However, our results showed that exon skipping occurred due to the CRISPR/Cas9‐mediated alterations at the splice site of a plant gene. Moreover, we found the exon 2 removal during the splicing process in the Rbadh2^ΔG and Rbadh2^ΔAAG mutants caused shifting of reading frame in the processed mRNA, resulting in a PTC in the exon 3 (Figure 1g). The existence of the PTC did not cause any further exon skipping in Rbadh2^ΔG and Rbadh2^ΔAAG, which was in line with the results obtained by Lee et al. (2020). The transcripts containing a PTC will be degraded by non‐sense‐mediated decay (Capito et al., 2018), as indicated by the reduced expression level of the mutated Osbadh2 compared with the WT (Figure 1d, h).

Since loss of OsBADH2 function promotes accumulation of 2AP, we explored the consequence of Osbadh2 exon 2 skipping on the 2AP content in grains of the homozygous transgene‐free plants of Rbadh2^ΔG and Rbadh2^ΔAAG using gas chromatography–mass spectrometry (GC‐MS). The Chinese fragrant rice variety Daohuaxiang No.2 (DHX2), which harbours a fragrant allele with mutation at exon 7 of OsBADH2, was utilized as a positive control. The GC‐MS internal standard for 2AP measurement was 2,4,6‐trimethyl pyridine (TMP) (Laohakunjit and Kerdchoechuen, 2007), as it has similar chemical properties to 2AP. Results showed that the 2AP content in the mutants (Rbadh2^ΔG and Rbadh2^ΔAAG) is as high as that in the positive control (about 0.08 mg/kg), while the 2AP in the WT was null (Figure 1i). These results indicate that the Osbadh2 in the mutants is not functional to hinder the production of 2AP, resulting in the grain fragrance.

We further investigated the effect of the OsBADH2 exon 2 skipping on the phenotypic characteristics of the mutants. Data showed that there were no significant differences between the edited plants and WT for vegetative and yield characteristics, indicating that the edited gene has no effect on the agronomic traits except the 2AP content or grain fragrance.

In conclusion, this study provides the first evidence that CRISPR/Cas9‐mediated deletion of the exonic nucleotide at the exon–intron junction of a plant gene could cause exon skipping during pre‐mRNA splicing. Furthermore, the OsBADH2 exon 2 skipping caused shifting in the reading frame, resulting in a downstream PTC in exon 3. Moreover, our results highlighted that CRISPR/Cas9‐mediated exon skipping could facilitate improvement of agronomically important trait of plants.