Transgenic manipulation of triacylglycerol biosynthetic enzymes in B. napus alters lipid-associated gene expression and lipid metabolism

Oilseed rape (Brassica napus) is an important crop that is cultivated for the oil (mainly triacylglycerol; TAG) it produces in its seeds. TAG synthesis is controlled mainly by key enzymes in the Kennedy pathway, such as glycerol 3-phosphate acyltransferase (GPAT), lysophosphatidate acyltransferase (LPAT) and diacylglycerol acyltransferase (DGAT) but can also be produced from phosphoglycerides such as phosphatidylcholine (PC) by the activity of the enzyme phospholipid: diacylglycerol acyltransferase (PDAT). To evaluate the potential for these enzymes to alter oil yields or composition, we analysed transgenic B. napus lines which overexpressed GPAT, LPAT or PDAT using heterologous transgenes from Arabidopsis and Nasturtium and examined lipid profiles and changes in gene expression in these lines compared to WT. Distinct changes in PC and TAG abundance and spatial distribution in embryonic tissues were observed in some of the transgenic lines, together with altered expression of genes involved generally in acyl-lipid metabolism. Overall our results show that up-regulation of these key enzymes differentially affects lipid composition and distribution as well as lipid-associated gene expression, providing important information which could be used to improve crop properties by metabolic engineering.


Results
Characterisation of B. napus transgenic lines during development. The three homozygous transgenic lines analysed in this study overexpressed transgenes encoding glycerol 3-phosphate acyltransferase 9 (GPAT9) from Arabidopsis thaliana (GPAT-OE), lysophosphatidic acid acyltransferase from Tropaeolum majus (LPAT-OE) and phospholipid: diacylglycerol acyltransferase1 from Arabidopsis thaliana (PDAT-OE). These represent the first two steps of the Kennedy pathway and the acyl-CoA-free formation of triacylglycerol (TAG), respectively. Transgene expression in each line was measured by RNA-seq analysis (Fig. 1A) and confirmed high-level expression as shown by normalised counts. A comparison of seed weight, embryo morphology and total fatty acid content in the transgenic lines in relation to the wild-type control at the mid-stage of development, which correlates with the phase of rapid oil accumulation, is shown in Fig. 1. While no significant alteration in embryo morphology was observed compared to WT (Fig. 1A), there was a small significant increase in fresh seed weight in PDAT-OE but not in GPAT-OE or LPAT-OE (Fig. 1B) at this developmental stage. Furthermore, both LPAT-OE and PDAT-OE accumulated significantly less total fatty acid per seed, with reductions of 12% and 36%, respectively, while there were no significant alterations in the GPAT-OE (Fig. 1C).

MALDI-MS analysis of PC and TAG molecular species quantification and spatial distribution.
Our observations indicated some changes in fatty acid and hence acyl lipid metabolism, therefore we compared seeds of the three transgenic lines with controls using MALDI-MS imaging to visualise the spatial distributions of phosphatidylcholine (PC) and triacylglycerol (TAG) molecular species in mature embryos. We had employed this method in B. napus previously to examine lipid metabolism during seed development 31 . Molecular species of PC were measured because this phosphoglyceride is particularly important as a metabolic intermediate during oil accumulation 27 .
There was very little change in the relative abundance of the predominant PC species in the GPAT-OE line compared to WT, except for an increase in 36:1 content from 1 to 3% and a small decrease in 36:2. However, in LPAT-OE and PDAT-OE lines there was an increase in the relative abundance of PC 36:2 and a small decrease in the amount of 36:3 and 36:4, together with a minor increase in 34:1. TAG species 54:3 showed little change in GPAT-OE and LPAT-OE samples but showed an increase in PDAT-OE (though this failed to meet the threshold for statistical significance due to the relatively large standard deviation in the WT sample) together with a small decrease in 54:4, 54:5 and 54:6 levels in PDAT-OE relative to WT.  Supplementary Fig. S1 and Fig. S2). The percentage of each molecular species was relatively quantified in the embryonic axis (hypocotyl, shoot and root poles) and the inner-and outercotyledons of WT embryos and those of the three transgenic lines and is shown in the bar charts of Figs. 2B and 3B and represented by heat-map colour intensity (green low, red high) on the MALDI-MS images. The WT and GPAT-OE MALDI-MS images for PC distribution were similar ( Fig. 2A,B). For example, the 34:2 species was predominantly concentrated in the embryonic axis in both of these genotypes, whereas the 36:2 species was lower in this region of the embryo relative to the cotyledons. Such a distinct heterologous distribution was not apparent in the LPAT-OE or PDAT-OE lines, with lower levels of 34:2, 36:3 and 36:4 and higher levels of 36:2 in the embryonic axis of these lines leading to a more homogenous distribution throughout the embryo.
For TAG, the MALDI-MS image data for WT and GPAT-OE were again broadly similar (Fig. 3A,B), with 52:2, 52:3 and 52:4 species showing predominant accumulation in the embryonic axis compared to cotyledons, while 54:4 and 54:5 were detected at higher levels throughout the embryo. TAG 54:3 showed homogenous accumulation in the WT sample with similar levels in cotyledons in GPAT-OE. By contrast, LPAT-OE and PDAT-OE showed reduced accumulation of 52:4 and 52:3 in the embryonic axis compared to WT. Moreover, there was a strong increase in 52:2 and 54:3 species in PDAT-OE in the embryonic axis and cotyledons and a moderate increase in these species in the same tissues of LPAT-OE. For 54:4 and 54:5 species, there were lower relative levels in PDAT-OE and only minor reductions in LPAT-OE and GPAT-OE.
In summary, GPAT overexpression has little effect on PC and TAG molecular species distribution in the embryonic tissues, while LPAT and PDAT overexpression reduced the distinct heterologous distribution. Overall, these results show that overexpression of key enzymes for TAG biosynthesis can have substantial effects on the usual pattern of oil accumulation found in seeds on high-oleic B. napus.  Table S2 for full list of differentially expressed genes), with substantially more genes displaying up-regulation compared to down-regulation in all samples. We performed gene ontology (GO) enrichment analysis on the sets of DEGs and revealed enrichment of specific classes of molecular function in each of the transgenic lines. Notably, GPAT-OE lines showed enrichment of genes associated with oxidoreductase activity, phosphotransferase and kinase activity, PDAT-OE lines showed enrichment of oxidoreductase and phosphoribosyl transferase activity, while LPAT-OE lines showed enrichment of a broader range of molecular functions, including fatty-acyl transferase activity, lipid binding, oxidoreductase activity, polysaccharide metabolism, primary carbon metabolism, transferase and transporter activity ( Fig. 4C-E).
We then examined the DEGs for each transgenic line to identify genes associated with acyl-lipid metabolism ( Fig. 4B; Table 1). Among the 36 genes which showed differential expression in all three transgenic lines, two genes were identified with roles in acyl-lipid metabolism: a putative cysteine S-palmitoyltransferase (CPT), which is used for lipid modification of proteins, and TRIGALACTOSYLDIACYLGLYCEROL2 (TGD2), which is involved in lipid transfer from ER to chloroplast. For GPAT-OE and LPAAT-OE lines, malonyl-CoA decarboxylase (MCD) was notably up-regulated. For LPAT-OE lines, putative ABC transporters, various lipid transport proteins and   Fig. 4B).
Because of the importance of plant acyl-CoA binding proteins (ACBPs) during seed development and oil accumulation 32 , we examined ACBP expression in embryos from developing B. napus seeds in the three transgenic lines compared to WT (Supplementary Fig. S3). Previous studies on B. napus gene expression provided background information about the relative expression of the six ACBPs (ACBP1-ACBP6) and their isoforms 33 . www.nature.com/scientificreports/ No significant differential expression of ACBPs was detected by RNA-seq analysis in any of the transgenic lines, but there was a decrease in ACBP2 expression in the LPAT-OE line that just failed to meet the differential expression threshold, indicating possible reduced activity of ACBP2 in the LPAT-OE line.

Discussion
In this paper, we studied the effect of overexpression of genes encoding three important enzymes for triacylglycerol formation, and hence oil accumulation, in Brassica napus. GPAT catalyses the first step of the Kennedy pathway and expression of two heterologous GPAT genes in Arabidopsis has been shown previously to produce modest increases in oil content 14 . In addition, other studies 15,16 have shown that boosting the supply of GPAT's substrate, glycerol 3-phosphate, can increase TAG accumulation in oilseed rape. There are several isoforms of GPAT in plants and analysis of the gene family reveals that the GPAT9 clade is responsible for TAG biosynthesis and oil accumulation 12 , and these functions have been demonstrated unequivocally in Arabidopsis by two separate groups 10,11 . We overexpressed Arabidopsis GPAT9 in oilseed rape and found no effect on total oil accumulation or significant change in the distribution of PC or TAG molecular species as revealed by MALDI-MS. However, biochemical experiments of oil accumulation in B. napus suggested that, because glycerol 3-phosphate levels were low 8 , endogenous GPAT activity was normally sufficient not to constrain TAG formation appreciably. Hence a lack of an observed increase in oil accumulation in GPAT-OE in this study may be due to the fact that GPAT activity does not exert significant flux control on TAG production.
In contrast to GPAT-OE lines, overexpression of the second Kennedy pathway enzyme, LPAT, caused an increase in seed TAG despite its low intrinsic flux control coefficient 24 . Moreover, we show here that LPAT overexpression changed the cellular distribution of PC and TAG molecular species. The MALDI-MS analysis showed that there was a general decrease in the unsaturation of TAG and, notably, of PC molecular species. Similar to the LPAT transgenics, the PDAT transgenics also showed significant alterations in the cellular distribution of molecular species of PC and TAG as well as a decrease in the unsaturation of these lipids. Although we did not evaluate the reason for the reduced unsaturation, it would be consistent with increased PDAT activity allowing less time for PC-mediated desaturation 34 . We note that total fatty acid content was reduced in the LPAT-OE and PDAT-OE lines but not in GPAT-OE at the mid-stage of seed development, which is the phase of rapid oil accumulation. This may in turn result in the altered lipid accumulation and distribution observed later in development in mature embryos, as shown by MALDI-MS analysis.
For the three transgenic lines, transcriptomic analysis revealed that LPAT-OE had the most DEGs, followed by GPAT-OE with the fewest DEGs detected in PDAT-OE, with substantially more up-regulated genes detected than down-regulated genes in all three transgenic lines. In oilseed rape, studies have shown that the final steps for TAG biosynthesis are catalysed mainly by the Kennedy pathway enzyme diacylglycerol acyltransferase (DGAT) rather than PDAT 6,28 . In that regard, it is perhaps not surprising that up-regulation of other enzymes involved in the Kennedy pathway (GPAT and LPAT) caused more changes in the DEGs detected by transcriptomic analysis than PDAT-OE.
Although the core genes encoding enzymes for oil accumulation showed no significant changes in the transgenic lines, there were notable alterations in the expression of other genes related to acyl-lipid metabolism. Several of these DEGs were common between different transgenics. For example, cysteine S-palmitoyltransferase (CPT), TRIGALACTOSYLDIACYLGLYCEROL2 (TGD2), malonyl-CoA decarboxylase (MCD), several ABC transporter G family members (ABCG) and a number of wax ester synthase/DGATs (WSD/DGAT) showed common alterations. CPT catalyses protein acylation and several isomers have been detected in plants 35 . TGD2 is involved in lipid transport between the ER and the chloroplast, while MCD has well-known functions in regulating fatty acid metabolism in bacteria and mammals 36,37 where the decarboxylation reaction has been mainly examined as part of the polyketide synthase proteins 38 . ATP-binding cassette (ABC) transporters function in transmembrane transport and plant genomes encode a large number which have been divided into eight subfamilies. They can transport a variety of substrates, including many lipids 39 . WSD/DGAT is a bi-functional enzyme which has been reported in a wide variety of organisms 40 . Its primary function is for the formation of surface wax esters 41 though it has been suggested that it could be responsible for making small amounts of TAG 42 .
The other DEGs identified in GPAT-OE include a gene encoding 4-coumarate-CoA ligase, which plays an important function in the biosynthesis of plant secondary compounds 43 . Two lipase genes were upregulated (GLIP3, Lipase 3) of which the GDSL esterase/lipase 3 plays a role in the esterification of lutein 44 although its gene family is diverse with multifunctionality 45 . A second protein acyltransferase (PAT 13) was upregulated while acyl-CoA binding protein2 (ACBP2) was downregulated, though this expression change did not meet our strict DEG criteria.
LPAT-OE showed additional DEGs compared to the GPAT-OE line. In particular, there was an increase in the expression of LEC1, which encodes a transcription factor that promotes expression of several genes involved in fatty acid biosynthesis 46 . In addition, there were more ABC G family transporter genes and WSD/DGAT genes whose expression was altered. In addition, several DEGs encoded lipid transport proteins (LTPs), which are encoded by multigene families and have functions in many physiological processes 47,48 . A total of 63 putative non-specific LTP genes have been identified in B. napus 49 . Other DEGs which have functions in lipid transport are seipin-1 50,51 and TRIGALACTOSYLDIACYLGLYCEROL2 (TGD2) 52 , the latter of which showed up-regulation in all transgenic lines.

Conclusions
Transgenic B. napus lines in which genes for key enzymes for TAG biosynthesis have been up-regulated showed differences in lipid accumulation, quality and distribution of PC and TAG molecular species, in addition to differential expression of genes involved in acyl-lipid metabolism and other molecular functions. These alterations could not be easily predicted based on our current knowledge of this important oil crop and emphasise the need for continuing work in the area. The observations reported here can form the foundation for such research.

Materials and methods
Plant materials. Brassica napus (oilseed rape) cv. DH12075 originally obtained from Agriculture and Agri-Food Canada, Saskatoon, Saskatchewan, Canada and homozygous LPAT-OE line TF2-14-85, GPAT9-OE line TF4-37-74, and PDAT1-OE line TF8-24-18 were provided by University of Durham. All transgenic lines contained single transgene insertions. These lines are referred to as WT, LPAT-OE, GPAT9-OE and PDAT1-OE in this work. Wild-type and transgenic seeds were germinated in pots containing soil/sand mix. Ten-day-old seedlings were transplanted individually into 8.7-inch pots. For data in Fig. 1, plants were grown in a greenhouse with a temperature of 23 °C and a light period (11-13 h)  The Arabidopsis thaliana PDAT1 sequence, At5g13640, was synthesized by Genscript, USA and cloned into pEntr1A-Napin-NosT between the Brassica napus Napin promoter and nos terminator sequences via NdeI and EcoRI sites. The Napin promoter-PDAT cassette was then transferred by Gateway LR Clonase II recombination into binary destination vector pRMH009, which contained the nptII gene expressed from the nos promoter, for selection of transformed material in tissue culture. The resultant final plasmid, pEW227-PDAT1, was verified by restriction mapping and sequencing the insert. The Arabidopsis thaliana GPAT9 sequence, At5g60620, was cloned into pEntr1A-Napin-NosT between the Brassica napus Napin promoter and Nos terminator sequences via NdeI and EcoRI sites. The Napin promoter-GPAT cassette was then transferred by Gateway LR Clonase II recombination into binary destination vector pRMH009, which contained the nptII gene expressed from the Nos promoter, for selection of transformed material in tissue culture. The resultant final plasmid was verified by restriction mapping and sequencing the insert. The construct used to produce LPAT/LPAAT (TmLPAT2) overexpressing lines of B. napus has been described previously 24 .
For PDAT-OE, twenty-four independent T0 transgenic plants were regenerated from one transformation experiment and confirmed as transformed by PCR amplification of the PDAT gene plus T-DNA copy number determination by Southern blotting. Three homozygous lines containing a single copy of the transgene were characterized and one representative line (line TF8-24-18) was used for subsequent analyses. For LPAT-OE, twenty-two T0 transgenic plants were regenerated from three independent transformation experiments and confirmed as transformed by PCR presence of the LPAAT gene of interest. The T-DNA copy number was determined by qPCR assay, based on the nptII gene and the number of T-DNA integration loci were determined by Southern blotting. Three homozygous lines with independent single-copy insertions of the transgene were identified and characterized, and one representative line (TF2-14-85) was used for subsequent analyses. For GPAT-OE, twentysix independent T0 transgenic lines were generated and confirmed as transformed by PCR amplification of the AtGPAT9 gene plus T-DNA copy number determination by Southern blotting. Three homozygous lines with independent single-copy insertions of the transgene were identified and characterized, and one representative line (TF4-37-74) was used for subsequent analyses.
Oilseed rape transformation was carried out as previously described 53 with Agrobacterium tumefaciens AglI pEW227-PDAT1, AglI pBI121/NastLPAT, AglI pRMH009/GPAT9 transformation of DH12075 cotyledonary petioles. Transformed shoots were regenerated in the presence of 15 mg l −1 kanamycin. Rooted shoots were transferred to Jiffy-7 pellets (Jiffy Products, Kristiansand, Norway), acclimatised to growth chamber conditions and transferred to 12 cm pots containing Levington M2 compost with 5 g/l slow-release fertilizer as above. T 0 plants were self-fertilised and T 1 seed harvested from all lines.
All plant research, including the use of transgenic plants, was conducted in compliance with international and UK guidelines.
Morphological analysis of Hong Kong-grown Brassica embryos at middle developmental stage. Due to the differences in light intensity and daylight length between the greenhouses in Hong Kong (11)(12)(13) h of natural light) and Cardiff (16 h with a light intensity of 250 µmol m −2 s −1 ) 28,31 , both wild-type and transgenic B. napus grew slower in Hong Kong than Cardiff. For Cardiff-grown wild-type B. napus, siliques were collected at 27 DAF, representing the mid-stage of lipid accumulation 28,31,54 ). For Hong Kong-grown B. napus, 38 DAF was selected to represent the middle stage of lipid accumulation. The Hong Kong-grown wild-type and transgenic embryos at 38 DAF were dissected manually by a razor blade and photographed 31 . Seeds from ten siliques from each of six plants were harvested for measurements of mature seed weight.
Fatty acid profiling. Twelve seeds from different siliques were collected at 38 DAF for each biological repeat. In total, six biological repeats were used. Fatty acid extraction was conducted as previously described 28 www.nature.com/scientificreports/ decanoic acid (19:0) (Sigma) (20 µg) was added as an internal standard. 2 ml of 5% NaCl and 3 ml of hexane were mixed with the samples, and the upper phase was collected after brief centrifugation. Subsequently, another 3 ml of hexane was mixed with the samples, and the upper phase was collected. The combined hexane phase was dried by nitrogen gas. FAMEs were analysed by an Agilent GC-MS device (5973 inert mass spectrometer combined with 6890N gas chromatograph) equipped with an Agilent J&W DA-WAX capillary column (30 m × 0.25 mm × 0.25 µm). The oven temperature was 170 °C for 3 min, increased at 4 °C min −1 to 220 °C, and held for 15 min at 220 °C 31 . FAMEs were identified by comparing the retention time of peaks with the Supelco 37 Component FAME MIX standard (Sigma).
MALDI-MS analysis. Dry seeds of B. napus were embedded in 10% gelatin, frozen for 16 h at − 80 °C and processed further as detailed 55 . Tissue sections (10 μm thickness) were imaged in a MALDI-LTQ-Orbitrap mass spectrometer (ThermoScientific). Set parameters were 12 μJ laser energy, 10 laser shots, 40 μm raster step size, collected over a m/z range of 700-1000 with the Orbitrap mass analyser set to a resolution of 60,000. Analysis used MS imaging software Metabolite Imager 56 and MSiReader 57 . Molecular species of TAG were collected as [M + K]+ adducts while those of PC were imaged as the sum of [M + H]= and [M + K]+ adduct intensities. The molecular classes used a mass tolerance of 10 ppm for imaging and are presented as mol% on a colour scale from green (low) to red (high). Calculation of average mol% for individual tissues summed the ion intensities of individual molecular species for each pixel where these represented laser spots and then normalising the total summed ion count for the PC or TAG lipid classes. Tissue areas evaluated were the embryonic axis, inner and outer cotyledons. Three replicate MS images were averaged and standard deviations plotted in bar charts for each genotype and the three tissue areas.

RNA extraction.
Siliques were harvested at 27 DAF. Embryos were manually dissected by a razor blade and stored immediately in liquid nitrogen before RNA extraction. Total RNA from embryos was extracted using RNeasy Plant Mini kit (Qiagen). On-column DNase digestion (Qiagen) was conducted on RNA samples to remove potential DNA contamination.
RNA-seq analysis. Embryo RNA samples from selected three transgenic lines (GPAT9-OE, LPAT-OE and PDAT1-OE) and wild type with three biological replicates, were used for transcriptomics analysis on an Illumina NextSeq500 machine. Paired end sequencing (2 × 75 bp) strategy was used in the RNA seq analysis. The B. napus genome from 'Darmor-bzh'was used as a reference genome 58 RNAseq reads were quality filtered using Trimmomatic v. 0.38 59 and resulting sequences were mapped with STAR v. 2.7.3 60 to the Plants Ensembl reference Darmor-bzh genome v.5 58 . Optical duplicates were removed with Picard Tools (http:// broad insti tute. github. io/ picard/) and genomic features were recovered from the corresponding Darmor-bzh v.5 annotation. RNA-seq reads were counted using featureCounts 61 and further analysed for differential gene expression using SARTools 62 and DESeq2 63 . Gene orthology and gene set enrichment analysis (GSEA) were performed with g:Profiler 64 against all known Brassica napus genes in Ensembl v.102 using a g:SCS threshold of 0.05. GSEA results were semantically summarised using REVIGO 65 against the whole UniProt database.