Design of High‐Diacylglycerol and Lecithin Soybean Seed Oil Using GmPDATs and GmDGATs Knockout via a CRISPR‐Cas9 System

Glycine max (L.) Merr. (soybean) is an important source for vegetable oil production, used for biofuel synthesis and human and animal consumption. Diacylglycerols (DAGs) are minor natural components of soybean oil and have been extensively studied as a functional food ingredient due to their different acylglycerol structures and metabolic characteristics compared with triacylglycerols (TAGs) (Lee et al. 2020). Dietary DAGs have been reported to maintain energy expenditure and prevent body fat accumulation in healthy individuals (Anikisetty et al. 2018). Soybean oil also contains various bioactive compounds, including lecithin, which is predominated by phosphatidylcholine (PC), followed by phosphatidylethanolamine (PE) and phosphatidylinositol (PI) (Scholfield 1981). Soybean lecithin is predominantly employed in the pharmaceutical and food industry because of its various potential health benefits, such as improving cognitive function, boosting immunity, and relieving menopause symptoms (Higgins and Flicker 2003). For optimal nutritional needs and healthy dietary practices, soybean seed oil should have increased DAG and lecithin contents. In higher plants, TAG synthesis is primarily modulated by core enzymes in the Kennedy pathway, including diacylglycerol acyltransferase (DGAT, EC2.3.1.20). Furthermore, TAG can also be synthesised by PC under the catalytic activity of the enzyme phospholipid: diacylglycerol acyltransferase (PDAT, EC2.3.1.43) (Zhang et al. 2009). In plants, the following three distinct DGAT classes have been identified: DGAT1, DGAT2, and DGAT3. These DGATs play different roles in TAG metabolism, among which DGAT1 primarily regulates TAG biosynthesis during seed development (Roesler et al. 2016). The enzymatic activity of PDAT was first identified in microsomal preparations from Ricinus communis, Helianthus annuus, and Crepis palaestina, prepared to study the application of phospholipids as an acyl and DAG as a donor and acceptor for TAG biosynthesis, respectively (Dahlqvist et al. 2000). In this study, soybeans with high DAG and lecithin content were developed using CRISPR/Cas9 editing techniques. Initially, eight PDAT genes were identified in the soybean genome based on the sequence homology to AtPDAT1/2 (Figure S1, Table S1). Then, due to its elevated expression, GmPDAT1B/1D/2A/2B was selected to study its impact on seed development (Figure S3). Furthermore, among the three DGAT types identified in the soybean genome, DGAT1-family genes (GmDGAT1A/B/C) were selected (Figure S2, Table S2) as they are important regulators of oil accumulation (Roesler et al. 2016). Notably, GmDGAT1A/B/C were markedly upregulated in developing soybean seeds (Figure S3), consistent with DGAT1 expression patterns in other plants. Moreover, GmDGAT1A/B/C and GmPDAT2A/B were specifically observed in the endoplasmic reticulum (ER), whereas GmPDAT1B/D was found in the chloroplast (Figure 1a). To investigate their function, seven genes were expressed in a yeast mutant with a TAG biosynthesis defect, and all of them produced TAG (Figure S4). Further, a three single guide RNA (sgRNA)-expression CRISPR/Cas9 vector was established with sgRNA1 targeting DGAT1A/B/C, sgRNA2 targeting PDAT1B/D, and sgRNA3 targeting PDAT2A/B (Figure 1b). In total, 34 positive T0 lines were acquired after the Agrobacterium tumefaciens-induced soybean cultivar DN50 transformation, among which 11 (32.35%) had mutations at the sgRNA target sites (Figure S5). CRISPR mutants with deletions and insertions induced translational frameshifts and were observed at all target sites in the T0 and T1 generations (Tables S3 and S4); however, no mutation events were identified at any potential off-target sites (Table S5). For further comprehensive analyses, eight homozygous mutants with diverse mutation combinations in the target genes, including dgat1ab/pdat2ab (c6-6), dgat1c/pdat2a (c11-7), pdat1d2b (c16-1), dgat1abc/pdat1b2a (c18-1), dgat1bc/pdat1d2ab (c18-3), dgat1ab/pdat1b2a (c23-1), dgat1ab/pdat1bd (c23-3), and dgat1b/pdat2a (c24-8) (Figure 1c,d), were selected. The transcripts of GmDGATs and/or GmPDATs were much lower in mutants than those in wild-type (WT) (Figure 1e), potentially due to alterations in target mRNA secondary structure caused by editing-induced indels (Figure S6). Mutant seeds exhibited a significant reduction in oil content (8.7% in c16-1 to 33.2% in c23-3), alongside a 2.9% to 11.4% increase in protein content (Figure 1f). Similarly, oil body (OB) size was consistently smaller in mutants than in WT (Figure 1g,h). Consistent with these observations, analogous results in Arabidopsis indicate that GmDGATs and GmPDATs share overlapping, essential roles in oil accumulation (Zhang et al. 2009). Notably, no marked differences in visible developmental or growth phenotypes were observed between the mutants and WT (Table S6). All mutant lines maintained stable seed yields, with oil production either matching WT or exhibiting modest reductions (Figure S7). To investigate lipid metabolism alterations in these mutants, the lipid profiles of developing seeds were analysed using electrospray ionisation tandem mass spectrometry (ESI-MS/MS). All mutant lines exhibited substantially elevated DAG and PC contents relative to WT, reflected by higher levels of all detected molecular species, such as 16:0/18:2, 16:0/18:3, 18:1/18:2, 18:2/18:2, and 18:1/18:3 (Figure 1i, Tables S7 and S8). Among mutants, c23-1 (dgat1ab/pdat1b2a) showed the highest increases, with 10.0-fold and 7.56-fold increases in DAG and PC, respectively. It was followed by c23-3 (dgat1ab/pdat1bd), which had 8.7-fold and 7.32-fold increases in DAG and PC, respectively (Figure 1i). Moreover, the contents of other phospholipids (PI, PS, PG, and PE) were also substantially elevated in all mutants, with PI increasing by 2.08–2.95-fold, PS by 1.21–1.82-fold, PG by 1.09–2.37-fold, and PE by 8.90–10.93-fold (Figure S8, Table S8). Consistent with the oil content analysis, the mutants displayed a downward trend in TAG content to different degrees (Figure 1i, Table S9). Additionally, thin-layer chromatography (TLC) analysis was carried out to assess changes in lipid composition in mutants and WT seed oil (Figure 1j). In the WT, the TLC profiles showed relatively weaker DAG bands, whereas the mutant lines had substantially enhanced DAG bands (Figure 1k). Lipid profile changes in these mutants arise from the distinct yet complementary roles of DGATs and PDATs in TAG biosynthesis. Knocking out GmDGAT1A/B/C blocks key steps of the classic acyl-CoA-dependent TAG synthesis pathway, leading to DAG accumulation (Figure 1l). Conversely, knocking out GmPDAT1B/D/2A/B (e.g., c16-1) inhibits the alternative TAG synthesis pathway involving PC and other phospholipids, thereby increasing PC-based lecithin (Figure 1l). Notably, simultaneous knockout of DGATs and PDATs exerts a synergistic effect that amplifies these changes, with combined disruption of DGAT1A, DGAT1B, and PDAT1B (e.g., c23-1, c23-3) yielding the most significant increases in DAG and PC levels compared to other mutation patterns. This suggests that concurrent disruption of both pathways more efficiently redirects carbon flux toward DAG and PC accumulation rather than their conversion to TAG, highlighting the coordinated role of DGATs and PDATs in balancing lipid biosynthesis and partitioning (Zhang et al. 2009). This study demonstrates the feasibility of CRISPR/Cas9-mediated engineering of soybeans with enhanced nutritional quality and commercial value by increasing contents of DAG and lecithin. Y.Z. performed most of the experiments and data analyses. B.Z., H.T., Y.L., Y.C., and S.W. provided technical support. Y.Z. wrote the manuscript. Z.H., M.Y., and Q.C. revised the manuscript. All authors have read and approved the final manuscript. This study was financially supported by Biological Breeding–National Science and Technology Major Project (2023ZD0403101), National Natural Science Foundation of China (32501888, U23A201783, 32272093). The data that supports the findings of this study are available in the Supporting Information of this article. Figures S1–S8. Tables S1–S10. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.