The Arabidopsis dgat1 knock-out mutant, AS11, purchased from NASC (CS3861), was used as the host for BnDGAT1 functional analysis experiments and the same genetic background Col-0 wild type was used as a control. All A. thaliana plants in this study were grown on a soil mixture (PINDASTRUP PLUS PEAT substrate: vermiculite: pearlite = 3: 1 : 1, Pindstrup Mosebrug A/S, Fabriksvej 2, 8550 Ryomgaard, Denmark) in a growth room at 22°C, and a 16 h/8 h light/dark cycle at a fluorescent light intensity of 100 μmol m^-2 s^-1.

The expression patterns of BnDGAT1 homologs were examined in different germplasm sources. In order to investigate the relationship between oil content and BnDGAT1 expression patterns, 14 lines with high seed oil contents (>50%) and 20 lines with lower seed oil contents (<45%) were selected. According to our breeding records, these lines showed relatively stable seed oil contents after more than 5 years of breeding. Because of long-term breeding, there are a few lines with extremely low oil content in our germplasm collection. Line designations are listed in Supplementary Table S1. In this study, all lines were grown and sampled in open fields.

Twenty-eight DAP (days after pollination) developing seeds from all thirty-four B. napus lines and 21 DAP developing seeds from 11 B. napus lines, were sampled for RNA extraction (Perry and Harwood, 1993). Tissue samples of roots, stems, rosette leaves, and the third cauline leaves of four B. napus lines (94-228, 1721-1B, 1941 and Danza 875) were collected at 28DAP and additionally, sepals, petals, stamens and pistils were harvested from these four lines at flowering, for expression analyses of BnDGAT1s. Based on our initial results of BnDGAT1s expression profiling in developing seeds, 12 lines with various BnDGAT1s expression patterns were selected for further analysis. Rosette leaves and flowers of these 12 lines were sampled for tissue-specific expression. Mature seeds from all these lines were collected for measurements of oil content and fatty acid composition of TAGs.

DGAT1 protein sequences were acquired from GenBank according to the annotation of sequences similar to BnDGAT1s. A Neighbor-joining (NJ) tree was constructed by MEGA6.0 (Center for Evolutionary Medicine and Informatics, Tempe, AZ, United States). DGAT1 protein sequences of the four BnDGAT1s were aligned by DNAMAN.

The coding region of four duplicated BnDGAT1s were cloned into plant expression vector pK7WG2D (Invitrogen) and construct integrity was confirmed by sequencing. PCR primers are listed in Supplementary Table S2. The vectors were transformed into Agrobacterium tumefaciens strain GV3101 and then transformed into A. thaliana dgat1 mutant AS11. Plant transformation was performed using the floral dip method (Clough and Bent, 1998).

Three pairs of primers were designed to detect the expression of BnDGAT1-1, BnDGAT1-2 and the combined expression of BnDGAT1-3 and BnDGAT1-4. The primers are listed in Supplementary Table S2. All primers were verified by amplifying respective plasmids with the four BnDGAT1s homeologs (Supplementary Figure S4).

Total RNA was isolated from various plant organs using Trizol reagent following the manufacturer’s instructions (Invitrogen, Carlsbad, CA, United States) and was treated with RNase-free DNaseI (Qiagen, Hilden, Germany). Total RNA (2 μg) was reverse- transcribed into cDNA using the M-MLV RTase cDNA Synthesis Kit (Takara, Shiga, Japan). The cDNAs were then used as templates in 20 μl semi-quantitative-PCR reactions using homeolog-specific PCR primers. Actin was used as the control gene. The semi-quantitative-PCR reactions for BnDGAT1-1, 2 and 3/4 were carried out under similar conditions (95°C for 30 s; 48–53°C for 1 min; 72°C for 30 s; 29 cycles) with a slight adjustment in annealing temperature (53°C for BnDGAT1-1, 53°C for BnDGAT1-2, and 48°C for BnDGAT1-3/4). Seven microliters of each PCR product was loaded onto a 2% agarose gel for electrophoresis.

Fatty acid analysis of Arabidopsis seeds followed the method of Li et al. (2006) with minor modifications. Thirty randomly-picked seeds per replicate were used for measuring seed weights after 48 h desiccation. Ten micrograms of seeds per replicate were used for fatty acid analysis. Triheptadecanoin was used as an internal TAG standard. Two milliliters of methanolic H₂SO₄ 2.5% (v/v) was added to each sample and heated at 80°C for 120 min. After neutralization and extraction with hexane, the fatty acid methyl ester profile was analyzed by gas chromatography on a DB23 column with flame ionization detection. Fatty acid composition analyses of B. napus seeds followed the same protocol except the seeds were dried at 60°C for 3 days and were pulverized before transmethylation.

Oil contents of B. napus seeds were measured on an mq20 NMR analyzer, Bruker optics (10 mm sample tube diameter, probe head temperature 40°C, environment temperature 25°C). Scans were repeated four times.

In previous studies, the four BnDGAT1s were heterologously expressed in yeast and found to have different acyl-CoA substrate preferences and varying efficiencies in catalyzing TAG biosynthesis (Aznar-Moreno et al., 2015; Greer et al., 2015). Protein sequence alignment comparisons among BnDGAT1s and the A. thaliana DGAT1, revealed eight known signature motifs, five of which were completely identical among the four homoeologs (Kaup et al., 2002), including an acyl-CoA binding signature spanning active site catalytic residues (motif A), a thiolase acyl-enzyme intermediate binding motif (motif B), a leucine zipper motif (motif D), a motif unique to DGATs (motif F), and a fatty acid binding protein motif (motif G) (Supplementary Figure S3). Residue differences were found on motifs C, E, and H (Figure 1). However, only one of them appeared in a structurally-important motif, on the first residue in motif C. Motif C (L¹⁸⁰-V¹⁸¹-X-R¹⁸³-X-X-X-S¹⁸⁷-X-X-X-A¹⁹¹) is a typical targeting site of members of the sucrose non-fermenting (SNF)-related protein kinase 1 (SnRK1) family, which are involved in the global regulation of carbon metabolism (Halford and Hardie, 1998). Similar SnRK1 targeting motifs were present in LPAATs from coconut (Knutzon et al., 1999) and meadowfoam (Lassner et al., 1995), and one was previously mentioned in the Arabidopsis DGAT1 (Zou et al., 1999). The first residue of motif C is methionine (M) in BnDGAT1-1 and BnDGAT1-2 while it is leucine (L) in BnDGAT1-3 and BnDGAT1-4. Since this residue is of structural importance, this difference might cause a functional divergence among BnDGAT1s. Other residue differences, including the two on motif C and five on motif E (a motif common to most annotated DGAT1s) and motif H (a DAG/phorbol ester binding motif) were variable residues among other DGAT1s as well.

To understand the evolutionary relationship of BnDGAT1s, 64 DGAT1s from 37 species were identified in GenBank based on the annotation of sequences similar to BnDGAT1s. A phylogenetic tree based on these protein sequences was constructed (Supplementary Figure S1). BnDGAT1-1 and BnDGAT1-3 displayed high sequence homology to DGAT1 genes of B. rapa (AA genome), while BnDGAT1-2 and BnDGAT1-4 were homologous to DGAT1 genes identified in B. oleracea (CC genome). Additionally, the chromosomal locations of the four BnDGAT1s were investigated according to genomic sequencing data on-line¹. However, there are only three BnDGAT1 genes in this database; one of them shows partial sequence identity to BnDGAT1-1 and partial sequence identity to BnDGAT1-2, suggesting an error in sequence assembly. Only BnDGAT1-3 could be assigned, located on chromosome A07 (40438–43985 bp) (Supplementary Figure S2).

In order to characterize activities and acyl-CoA substrate preferences of the four BnDGAT1s in planta, their coding regions were expressed individually in AS11, an Arabidopsis dgat1 knock out line in the Col-0 genetic background (Katavic et al., 1995). Loss-of-function of DGAT1, as is the case in AS11, leads to decrease of seed oil content and drastically alters its fatty acid profile. To avoid possible complications in interpretation of results due to dosage effects of multiple insertions (Vassileva et al., 2001; Mansur et al., 2005), single-copy transgenic lines were identified by the segregation ratio of T₂ seeds. Ten of these single insertion lines were randomly selected and characterized; seed oil contents and fatty acid compositions were determined by GC. The results showed that heterologous-expression of any one of the BnDGAT1s restored the oil deficiency in AS11 to the oil level found in the wild type (Figure 2A). Additionally, all DGAT1s generally complemented the AS11 fatty acid composition changes, restoring them to a composition more like wild type. Three FAs (18:1, 18:3, and 20:1) showed significant reversion to WT proportions in all 40 homozygous BnDGAT1 AS11 transformants, when compared to AS11. Eicosenoic acid (20:1) increased, and 18:3 decreased to the level of wild type (Figures 2B–D). In general, the total proportion of mono-unsaturated fatty acids was restored. These results suggested that all of the four expressed BnDGAT1s had DGAT enzyme activity.

Because all four BnDGAT1s were expressed in the same AS11 background without the interference of endogenous/native AtDGAT1 activity, phenotypes of these transgenic lines, to a large extent, reflect activities of the expressed BnDGAT1s. It is worth noting that the four BnDGAT1 transformant sets did not generally show widely divergent differences in oil accumulation and fatty acid composition, which also suggests that they have similar kinetics and acyl-CoA preferences in planta.

Although the heterologously expressed cDNAs of BnDGAT1s did not show differences in activity (as measured by oil content), there might be differences in gene expression. In a previous study, expression levels of BnDGAT1-1 and BnDGAT1-2 were much higher than those of BnDGAT1-3 and BnDGAT1-4 in middle- and late-stage embryogenesis (Aznar-Moreno et al., 2015). We were interested in whether this would remain true in other germplasm sources. In other words, are all four DGAT1s crucial to TAG synthesis in B. napus? In order to address these questions, characterization of BnDGAT1s expression patterns in different germplasms was undertaken. However, the very high sequence similarity/homology of the four genes renders it difficult to discriminate between them by traditional real-time PCR or northern-blot. Although the expression patterns of the four genes in B. napus were identified by transcriptome sequencing (Aznar-Moreno et al., 2015), it would be intractable and costly to quantitatively test dozens of accession lines and tissues by RNA-Seq. Therefore, as a compromise we selected semi-quantitative PCR (semi-q PCR) for testing the expression difference of the four BnDGAT1s. Even so, BnDGAT1-3 and BnDGAT1-4 were still too similar to be distinguishable from each other; thus one pair of semi-q PCR primers was designed to measure the combined expression of BnDGAT1-3 and BnDGAT1-4. The other two pairs of semi-q PCR primers were designed to measure the expression of BnDGAT1-1 and BnDGAT1-2, independently. Semi-q PCR measurements of expression were carried on 28 DAP developing seeds of 34 inbred lines and 21 DAP developing seeds of 11 inbred lines. The results showed that there were different expression patterns in this germplasm collection. Three/four BnDGAT1s distinctly expressed in 13 lines and all four genes weakly expressed in two lines. Not all BnDGAT1s strongly expressed in the other 19 lines (Supplementary Figure S5). It is worth noting that the expression patterns are identical in the same germplasm at two stages of developing seeds. Taken in the context of the above results of expression in Arabidopsis, and AS11 in particular, all four BnDGAT1 are likely functionally-expressed in different germplasms.

Based on strong semi-q PCR band representation of the BnDGAT1s identified, there were eight types of expression patterns (Figure 3A) and the 34 lines fell into different expression types of BnDGAT1s (Supplementary Table S3). The relationship between BnDGAT1 expression patterns and seed oil contents was studied in all 34 rapeseed lines (Supplementary Table S1). Lines with lower seed oil contents primarily have one or two weaker-expressed or silenced BnDGAT1s, examples being lines 1721-1B, QY3, You22 (Supplementary Figure S5); some even have three or four weaker-expressed or silenced BnDGAT1s, as in lines in groups III and IV (Figure 3). However, this trend was not consistent in all lines. Some lines with the expression of four BnDGAT1s, such as QY10 (Supplementary Figure S5), still showed low seed oil content. To further investigate the relationship between seed oil contents and expression patterns of BnDGAT1s, average oil contents of each pattern type were calculated. Oil content of type A with three strong semi-q PCR bands was much higher than those of other types (Figure 3B). Type B and C showed slightly higher seeds oil contents than the other groups. Considering the similar functions of each DGAT1, lines were grouped by the number of distinct DGAT1 expression bands. There were two sharp bands in types B, C, and D and they were designated as Group II. Group III included types E, F, and G with one sharp band. Group I and Group IV included lines with three sharp or weak bands, respectively (Figure 3A and Supplementary Table S3). Average oil contents of groups were calculated and results showed that the higher the expression of BnDGAT1s, the higher the oil contents of the groups (Figure 3C). These results suggest that seed oil content of the rapeseed lines was closely correlated to the expression level and pattern of BnDGAT1s.

BnDGAT1s were reported to show diversity in substrate specificity (Aznar-Moreno et al., 2015). In order to investigate the possible consequence of this difference in planta, fatty acid compositions were compared among type G (only BnDGAT1-1 highly expressed), type F (only BnDGAT1-2 highly expressed) and type H (none BnDGAT1s highly expressed). The proportions of 18:1 were 50–60% in all of these lines and no obvious difference in fatty acid composition was found among these lines (Supplementary Figure S6).

To determine whether there are different tissue expression patterns of the four BnDGAT1s within the same germplasm background, lines with different BnDGAT1 expression patterns in developing seeds (Danza875 type A from group I, 1721-1B type B from group II, 94–228 type G from group III and 1941 type H from group IV), were investigated. Roots, stems, the first rosette leaves, the third cauline leaves, sepals, petals, stamens, and pistils and developing seeds (28 DAP) were collected at flowering. Expression analyses of BnDGAT1s in these tissues showed that BnDGAT1s mainly expressed in reproductive organs. Stamens were the organ with the strongest expression of the BnDGAT1s, followed by pistils, 28 DAP seeds and petals. There was no detectable expression of BnDGAT1s in roots, stems, and rosette leaves at this stage. However, it was interesting to note that there was weak expression of BnDGAT1s in cauline leaves (Figure 4). Other than the expression level, the expression patterns of the four BnDGAT1s in different organs were identical to that in developing seeds (Figure 4). In order to test the similarity of BnDGAT1 expression pattern in other lines, samples of leaves, flowers, and developing seeds were collected from 12 lines with different DGAT1 expression patterns. The BnDGAT1 expression pattern in flowers was identical to that in developing seeds in all 12 lines and no exceptions were found in any of the 12 lines that were tested. BnDGAT1s that were not expressed in developing seeds, like BnDGAT1-2 in QY3, B120, 1886, were also not expressed in other organs (Figure 5). These genes may be silenced in these lines. The data suggest that the four BnDGAT1s have consistent expression patterns in different tissues, especially in flowers and developing seeds, spanning the generative phase.