Overexpression of a type 2 diacylglycerol acyltransferase in the marine diatom Phaeodactylum tricornutum enhances lipid production and omega-3 Long Chain Polyunsaturated Fatty Acid content CURRENT STATUS: UNDER REVIEW

Background: Oleaginous microalgae represent a valuable resource for the production of high value molecules. Considering the importance of omega-3 long chain polyunsaturated fatty acids (LC-PUFAs) for human health and nutrition the yields of high value eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) require significant improvement to meet demand, however, the current cost of production remains high. A promising approach is to metabolically engineer strains with enhanced levels of triacylglycerols (TAGs) enriched in EPA and DHA. Results: Recently we have engineered the marine diatom Phaeodactylum tricornutum to accumulate enhanced levels of DHA in TAG. To further improve the incorporation of omega-3 LC-PUFAs in TAG we focused our effort on the identification of a type 2 acyl-CoA:diacylglycerol acyltransferase (DGAT) capable of improving lipid production and the incorporation of DHA in TAG. DGAT is a key enzyme in lipid synthesis. Following a diatom based in vivo screen of candidate DGATs, a native P. tricornutum DGAT2B was taken forward for detailed characterisation. Expression of DGAT2B in P. tricornutum was confirmed by qRT-PCR and the transgenic strain grew successfully in comparison to wildtype. PtDGAT2B has broad substrate specificity with preferences for C16 and LC-PUFAs acyl groups. Moreover, the expression of DGAT2B resulted in higher lipid yields and enhanced levels of DHA in TAG. Furthermore, co-expression with a Δ5-elongase showed how iterative metabolic engineering can be used to increase DHA and TAG content, irrespective of nitrogen treatment. Conclusion: This study provides further insight into lipid metabolism in P. tricornutum and suggests a metabolic engineering approach for the efficient production of EPA and DHA in microalgae.

3 microalgae, however their negative effect on cell growth is a major bottleneck for reducing production costs. Therefore, new efficient strategies are needed to increase economic viability of production [5].
A promising approach is the use of genetic engineering for generating improved strains with desirable characteristics such as fatty acid (FA) composition, enhanced lipid biosynthesis and high growth rates.
Within the complex network of activities associated with TAG assembly, acyl-CoA: diacylglycerol acyltransferases (DGATs) and phospholipid:diacylglcerol acyltransferases (PDATs) have been shown to play an important role in TAG accumulation (for reviews see [1] and [6]). DGAT (EC 2.3.1. 20) catalyses the final committed reaction in the acyl-CoA dependent Kennedy TAG biosynthetic pathway by the esterification of acyl-CoA to a diacylglycerol (DAG) moiety [7,1]. At least three classes of DGATs have been identified in eukaryotes. Of these, two types, designated DGAT1 and DGAT2, are membrane-bound acyltransferases and play a major role in TAG biosynthesis. Whereas DGAT3 is a soluble isoform characterised in peanut (Arachis hypogaea) and Arabidopsis thaliana, where it is thought to regulate the flux between the cytosolic acyl-CoA pool and TAG [1, 8,9]. The pivotal role of DGAT in TAG synthesis has been demonstrated by the increase in seed oil content and altered FA composition resulting from the overexpression of plant DGAT1 and DGAT2s (for Refs see [1,10,11]). Therefore, one beneficial route for the metabolic engineering of microalgae for increased oil content and tailored FA composition may be the manipulation of microalgal DGAT activities.
Most eukaryotic organisms contain at least one copy from the DGAT family of three classes [1]. To date, genome sequencing of photosynthetic microalgae has revealed the presence of a single gene copy encoding the type 1 DGAT, apart from Nannochloropsis oceanica IMET1 where two isoforms were identified [6]. By contrast, multiple copies of DGAT2 genes have been annotated in microalgal genomes, for example five DGAT2s have been found in Phaeodactylum tricornutum and 12 copies in N. oceanica [12,13]. Despite this diversity, currently only a few microalgal DGATs have been functionally characterized. In most instances attempts to functionally characterise DGAT isoforms have used a Saccharomyces cerevisiae mutant strain (H1246) lacking endogenous DGAT and PDAT genes. Using this approach DGAT1 isoform from P. tricornutum was successfully characterised, demonstrating a preference to produce TAGs with a high level of saturated fatty acids (16:0 and 18:0) 4 [12]. Numerous DGAT2 genes have been putatively identified in microalgae e.g. N. oceanica, Chlamydomonas reinhardtii, Ostreococcus tauri, Thalassiosira pseudonana and P. tricornutum, for subsequent characterisation in TAG deficient S. cerversiae [14,15,16,17,18,19]. Not all the genes tested restored the TAG phenotype, but of those that did, many showed differing substrate specificity for mono-and polyunsaturated acyl-CoA substrates, e.g. O. tauri DGAT2B showed broad substrate specificity, accepting saturated as well as mono-and polyunsaturated acyl-CoAs as substrates [18], while DGAT3 isoform from P. tricornutum, (PtDGAT3) displayed a preference for the incorporation of unsaturated C18-fatty acids [20].
In P. tricornutum four putative type 2 DGATs, annotated as PtDGAT2A (Phatr2_49462), PtDGAT2B (Phatr2_49544), PtDGAT2C (Phatr2_31662) and PtDGAT2D (Phatr2_43469), have been characterised in TAG deficient yeast [19]. Of these only PtDGAT2B restored lipid body formation and showed a preferential substrate specificity for unsaturated FAs. Additionally, transcript analysis revealed that PtDGAT2A and PtDGAT2B were upregulated prior to the onset of TAG accumulation. Further studies indicated that the overexpression of PtDGAT2D in P. tricornutum resulted in an increased flux of photosynthetically fixed carbon towards lipids leading to a two-fold higher lipid content in transgenic cells compared to wild type (WT), alongside a reduction (15%) in cell growth [21]. In contrast, overexpression of a chloroplast DGAT2 homolog from the diatom T. pseudonana, (Thaps3_25595), significantly increased TAG accumulation without a negative effect on growth or biomass accumulation [22].
Collectively, these studies illustrate how manipulation of DGAT genes in microalgae can increase their lipid content and modify fatty acid composition. Moreover, it is reasonable to suggest that they might participate in producing TAGs with enhanced levels of long-chain polyunsaturated fatty acids (LC-PUFAs) in microalgae (i.e. increasing incorporation of eicosapentaenoic acid, 20:5Δ 5,8,11,14,17 [23] and demonstrated the potential of this transgenic strain for industrial production of omega-3 LC-PUFAs in phototrophic and heterotrophic conditions [24,25]. The DHA content in transgenic Pt_OtElo5 was increased up to eight-fold that of WT levels and several novel LC-PUFA containing TAG species were detected. Further work has demonstrated the feasibility of multifunctionalizing the Pt_OtElo5 strain, combining the accumulation of high levels of EPA and DHA with recombinant protein production [26]. In the work presented here, we studied the effect of transgenic expression of the three most promising candidate DGAT2's, namely P. tricornutum DGAT2A, DGAT2B and T. pseudonana DGAT2 (TpDGAT2), on lipid accumulation and omega-3 LC-PUFA incorporation into TAG in WT and transgenic strains. The impact of these enzymes on lipid production in vivo was studied via lipidomic analysis. Our results demonstrate that overexpression of PtDGAT2B significantly enhances TAG production and alters FA composition during nitrogen-replete and nitrogen-limiting conditions.
This study provides better insight into TAG biosynthesis in P. tricornutum and suggests a promising approach for tailoring oil composition in microalgae.

Results
Generation of transgenic P. tricornutum strains overexpressing candidate DGAT2s.
The native codon sequences for P. tricornutum DGAT2A, DGAT2B and a T. pseudonana DGAT2 (TpDGAT2), codon-optimized for expression in P. tricornutum, were cloned into the pPhOS2 vector [23] and the resulting constructs (Pt_DGAT2A, Pt_DGAT2B and Pt_TpDGAT2) were then transformed into P. tricornutum via biolistics (as described in [23]). Multiple zeocin-resistant colonies obtained for each construct were confirmed by PCR and then used to inoculate cultures for further screening by gas chromatography coupled to ion flame detection (GC-FID) analysis of fatty acid methyl esters (FAMEs). From this screening, we obtained multiple independent transgenic lines for each construct (Additional file 1: Table S1). The FA profiles and lipid content of selected independent overexpression lines were then analysed and compared to that of WT. As previously observed [23] the main FAs in total lipid extracts of WT and transgenic cells in decreasing order were C16:1, EPA, C16:0 and C14:0 ( Fig. 1a). Transgenic lines overexpressing Pt_DGAT2B and Tp_DGAT2 demonstrated slightly elevated DHA levels.
To visualise the impact of transgene expression, a Nile Red fluorescence assay was used to evaluate 6 the ability of the candidate DGAT2 overexpressing lines to accumulate enhanced levels of neutral lipids. Previously, analysis of transcript levels of the DGAT2 genes under standard growth conditions showed that the expression level of PtDGAT2A and PtDGAT2B progressively increased from day two to day four, peaked and then decreased to a low level in the following days [27]. Hence, transgenic strains overexpressing DGAT2s were cultivated for three days in standard F/2 medium to the early exponential (E) phase (2.5x10 6 cells). Nile Red staining of neutral lipids showed substantial differences between different DGAT2 clones and WT (Fig. 1b). All PtDGAT2B expressing clones showed a marked increase of TAG, while Pt_DGAT2A and Tp_DGAT2 cells did not accumulate elevated oil levels compared to WT cells.
To confirm gene transcription and elevated expression of DGAT2 genes, quantitative real-time PCR (qRT-PCR) was carried out on selected transgenic lines. Based on the results of Nile Red staining, one of the most promising transgenic lines overexpressing each of the DGAT2 isoforms were characterized. Transgenic cultures were grown in F/2 medium and RNA extracted at the early E phase.
Subsequent qRT-PCR analysis confirmed expression of Pt_DGAT2A_1, Pt_DGAT2B_5 and Tp_DGAT2_6 transcripts in the selected transgenic strains and demonstrated that the Pt_DGAT2B transgene is expressed at higher levels than the other two transgenes, Tp_DGAT2 and Pt_DGAT2A (Fig. 1c).
Although the genes were under the control of the same promoter, the relative expression levels for the three DGAT2 isoforms were significantly different (p < 0.001, for expression of both DGAT2A and of DGAT2B genes). For cell lines overexpressing native PtDGAT2A and PtDGAT2B, transcripts were increased relative to WT by 2.3 and 3.0-fold, respectively. Based on the combined analysis of FA profiles, Nile Red staining and qRT-PCR analysis, the clone with the highest lipid content, Pt_DGAT2B_5, now designated DGAT2B, was taken forward for further detailed analysis. To evaluate the effect of PtDGAT2B overexpression on the incorporation of omega-3 LC-PUFAs in TAG, coding sequences for the O. tauri Δ5-elongase (OtElo5) and PtDGAT2B were cloned in the previously described pPhOS2 vector [23], generating DGATElo construct. Ten individual zeocin-resistant clones were confirmed by PCR and screened for FA content (Additional file 2: Table S2). Four of the independent clones containing the highest levels of the new fatty acid, docosapentaenoic acid (DPA, 22:5n-3), the product of elongating activity of Δ5-elongase, and DHA were selected for further analysis. FAMEs analysis of the selected transgenic strains confirmed the presence of DPA in the range of 1.8-4.6% accompanied by increased DHA levels (up to 3.8-8.5% compared to 1.8% in WT) in the stationary (S) phase of growth (Fig. 2a). Compared to WT, transgenic clones demonstrated substantially altered fatty acid profiles, containing on average 2-fold higher levels of 16:3 and 16:4, while the levels of C18:2 and C18:3n-3 were slightly reduced. The lipid content of these clones was further screened by Nile Red assay (Fig. 2b). The clone with the highest level of DHA and enhanced neutral lipid content, DGATElo_8 (designated DGATElo) was taken forward for further analysis. Nutrient deprivation has been shown to induce TAG accumulation in microalgae [4]. To address the role of PtDGAT2B in TAG synthesis we studied cell growth under N-replete and N-deplete conditions during the most active phase of oil accumulation (72 hrs). N-replete F/2 media was supplemented with additional P and N according to Abida et al., to ensure that these elements were not limiting during cell growth [28]. N-deplete media contained F/2 enriched with added P and no N. The growth rates and lipid accumulation of transgenic lines were determined under both N-replete and N-depleted conditions at 24-hour intervals for a total of 72 hours. The analysis of the growth profiles showed that there were no differences in growth patterns between the transgenic lines and the WT under both conditions. The rate of cell division under both N-replete and N-deplete conditions was both linear and consistent for transgenic and WT cells. (Fig. 3a, b). Whereas the overall level of cell division differed substantially between the two nitrogen conditions (p < 0.01), decreasing in the N-deprived cultures. To confirm transcription of both the OtElo5 and PtDGAT2B genes during the experiment, RNA was extracted from triplicate cultures at 72 h of growth after resuspension in either N-replete or N-free media. Following cDNA synthesis qRT-PCR confirmed that overexpression of PtDGAT2B and transcription of the OtElo5 gene were occurring under experimental conditions (Fig. 3c, d). We have 8 previously described the successful expression of OtElo5 under the endogenous fcpA promoter of highly expressed fcpA gene [23]. However, fcp promoters may not be strong enough to overexpress introduced activities due to the presence of light-responsive cis-regulatory elements [29] and stability under low nutrient conditions. Seo et al. showed that the endogenous elongation factor 2 (EF2) promoter isolated from P. tricornutum drove the expression of a transgene 1.2-fold stronger than that driven by the fcp promoter in light conditions and was stable throughout light and dark cycles [30].
However, recently we have shown that the expression of the Δ5-elongase (OtElo5) gene with an EF2 promoter resulted in comparable levels of EPA and DHA to that in transgenic strains in which the OtElo5 gene was under control of the fcpA promoter [26]. In this study we tested both promotors' efficacies in response to Nitrogen treatment in transgenic lines expressing OtElo5 (Additional file 3: Table S3). The results indicate that the fcpA promoter is suitable for transgene expression of DGAT2B in P. tricornutum when grown in different N conditions.

Effect of PtDGAT2B overexpression on fatty acid composition under N-replete and Ndeplete conditions.
To assess the impact of PtDGAT2B overexpression on cell FA profiles, cells were grown in N-replete and N-deplete conditions and analysed at 24, 48 and 72 h time points respectively. The fatty acid profiles of selected independent overexpression lines DGAT2B and DGATElo were then compared to that of WT and transgenic Pt_OtElo5. Fatty acid composition (Mol%) was clearly affected by N treatment, for example DGAT2B cells accumulated slightly higher levels of DHA in comparison to WT in N-replete conditions (Fig. 4). Irrespective of N treatment, the highest DHA levels were observed in DGATElo cells. As expected, cells expressing OtElo5 (Pt_OtElo5 and DGATElo) contained decreased EPA compared to that of DGAT2B and WT at all time points due to the action of the Δ5-elongase (Fig.   4). The novel FA DPA was only present in the strains expressing OtElo5 (Pt_OtElo5 and DGATElo).
When total FAMEs were quantified (expressed as nmol mg -1 ), the main differences between strains were observed in the content of 16:0 and 16:1 (Additional file 4: Figure S1). As expected, [28,31] Although PtDGAT2B has a strong substrate preference for C16 FAs, it also demonstrates the broad substrate specificity of this acyltransferase, particularly towards LC-PUFAs, such as EPA, DPA and DHA.

Lipidomic analysis of WT and transgenic cells under N-replete and N-deplete conditions
To gain insight into the ability of PtDGAT2B to enhance TAG accumulation we performed a The relative contribution each lipid has on the direction of each principle component was determined.
Additional file 5: Table S4 lists the 15 lipids with highest loadings associated with these dimensions.  Figure   S1 and Additional File 6: Figure S2). Introduction of the OtElo5 gene separates out cell lines expressing it under N-replete and N-deplete conditions. Cells expressing the OtElo5 gene and overexpressing PtDGAT2B, individually or in combination, cluster together. These differences between the cell lines are driven by lipids containing DPA and DHA, e.g. DGTA 44:12 among others. As expected, the PCA plots show that the biggest impact on distribution is N stress and that this is driven by changes in abundance of TAG. Other lipid species contribute to changes in lipid profile, but they are largely minor components. These data suggest that in response to N stress, de novo biosynthesis of TAG primarily utilizing the ER-localized Kennedy pathway, rather than remodelling of membrane lipids, is the major driver in TAG accumulation.

Impact of transgene expression on TAG accumulation under N-replete and N-deplete conditions
First, we examined the impact of PtDGAT2B overexpression on lipid accumulation in transgenic cells.
Both the DGAT2B and DGATElo cells contained elevated TAG compared to WT and Pt_OtElo5 cells irrespective of N conditions at all-time points (Fig. 6 and Additional file 7: Table S5). The results demonstrated that N deprivation is not required for enhanced TAG accumulation in the PtDGAT2B showed that along with the different levels of TAG accumulation between strains, there were also differences in TAG FA profiles and molecular species composition under N-replete and N-deplete conditions. DGATElo contained a significantly higher proportion of DHA in TAG than other strains under both conditions (p=0.0003) (Fig 7 a, b). However, DHA levels were slightly decreased in all cell types under N starvation in agreement with previous observations that LC-PUFAs levels decrease in Ndepleted medium [32].
Also, in agreement with previous reports [28,31]  The levels of DHA-containing lyso-PC were 2-fold higher in DGATElo compared to that of OtElo5 cells.
Overexpression of the DGAT2B gene on its own has only small impact on the DHA levels in N-replete cultures. These data indicate a crucial role lyso-PC in LC-PUFA synthesis and incorporation into complex lipid species.

Discussion
DGATs are thought to be rate-limiting enzymes and are considered the main contributors to oil biosynthesis in microalgae. Most microalgal species, including P. tricornutum have multiple copies of DGAT2 genes. Hence, identification of the best candidates for enhancing LC-PUFA production and TAG accumulation represents a promising approach to the metabolic engineering of strains with desired qualities. However, a functional role for the multiple type 2 DGATs found in algal species remains unclear. In microalgae N stress triggers TAG biosynthesis and often induces DGAT expression [6]. In P. tricornutum during N-replete and N-deplete conditions, TAGs and the precursor DAG are mainly composed of 16:0 and 16:1, whereas EPA and DHA are minor species [31,28]. This suggests that LC-PUFAs are likely transferred to TAGs by either one of the four DGAT2s identified in P. tricornutum [27] via the Kennedy pathway or a route involving phosphatidyl diacylglycerol acyltransferase PDAT [2].
We investigated the impact of the overexpression of DGAT2 isoforms in P. tricornutum WT and Pt_OtElo5 strains on overall TAG accumulation and specifically DHA partitioning in TAG. To achieve this, transgenic cell lines overexpressing the three most promising candidate DGAT2s, PtDGAT2A, PtDGAT2B and TpDGAT2, were generated and analysed for their fatty acid composition, relative overexpression levels and lipid content. Transgenic cells overexpressing PtDGAT2B showed the highest mRNA levels (a 3-fold increase when normalized to WT strain), elevated lipid content evaluated via a Nile Red fluorescence assay (Fig. 1) and increased proportions of DHA in total FAs (Fig. 2). To further improve LC-PUFA content and partitioning in TAG we coexpressed PtDGAT2B, the most efficient among the three studied DGAT2B candidates in enhancing neutral lipids levels, with the heterologous O. tauri Δ5-elongase (OtElo5) and analysed the FA profiles and lipid production in cells grown in N-replete and N-deplete medium.
FAMEs analysis indicated that PtDGAT2B can significantly alter FA composition in transgenic clones under N-replete and N-deplete conditions. The main differences between strains were observed in the levels of 16:0, 16:1, EPA, DPA and DHA (Fig. 3, Additional file 1: Table 1). In clones overexpressing  Table S4).

A 4-fold increase in TAG accumulation observed in DGAT2B cells under N-replete conditions
demonstrates that the requirement for nutrient depletion to increase lipid accumulation could be overcome by using PtDGAT2B as a tool for manipulating TAG content. Expression of the transgenes did not significantly alter cell growth rate in the different cell lines (Fig. 3). This contrasts with a previous study in which a PtDGAT2D gene was overexpressed in P. tricornutum; cells contained elevated TAG but grew 15% slower than WT [21].
From a biotechnology perspective, the ability to engineer strains that overproduce TAG under normal growth conditions is highly desirable as nutrient depletion negatively affects cell growth.
Overexpression of another type 2 DGAT, NoDGTT5, in N. oceanica CCMP1779 grown in N-replete medium resulted in increased TAG synthesis normally observed only after N deprivation [15].
However, lipid accumulation was accompanied by strongly reduced growth rates normally observed under N-deplete conditions. N starvation triggers substantial TAG accumulation in all cell types with the highest 37-fold increase in DGATElo cells when compared with N-replete conditions and 1.8-fold higher level of TAG than in the WT in N-deplete medium. Under N stress TAG levels in DGAT2B transformants increased 11-fold when compared with N-replete conditions and ~1.5-fold compared with these in WT cells (Fig. 7, Additional file 5: Table S4). Recent studies of DGAT isoforms demonstrated that in P. tricornutum only PtDGAT1 was strongly regulated by nitrogen depletion [14], whereas PtDGAT2B was highly expressed under Nreplete conditions before the onset of TAG accumulation [27]. Our findings suggest that PtDGAT2B might contribute to TAG accumulation under both N-replete and N-deplete conditions since we observed significant increase in TAG content in transgenic cells overexpressing PtDGAT2B irrespective of N treatment. We also demonstrated that the fcp promoter is suitable for gene overexpression under N deprivation (Fig 3.; Additional file 3: Table S3). This contrasts with the results of expression of a NoDGTT5 under the control of EF promoter in N. oceanica in which elevated TAG was observed under N-replete, but not deplete conditions. However, the use of the native gene promoter did increase TAG biosynthesis under deplete conditions [15].
Lipidomic analysis demonstrated that PtDGAT2B possesses broad substrate specificity and can incorporate different acyl-CoA species, including C16 and LC-PUFAs. Overexpression of DGAT2B resulted in an increase in TAGs containing C16 species, whereas cells expressing OtElo5 accumulated elevated levels of LC-PUFA containing TAG (Fig. 7). The levels of DHA-containing TAG species were higher in cells coexpressing both OtElo5 and DGAT2B indicating that PtDGAT2B enhances the incorporation and transfer of endogenous C16:0 and C16:1 fatty acids and newly synthesized DPA and DHA into TAG.
Changes in membrane lipids were principally observed in MGDG, DGDG and PG reflecting the rapid senescence of chloroplast membranes, whilst total amounts of PC, DGTA and lyso-PC did not change significantly over the time in all strains. Notably there were significant reductions in EPA-and C16containing MGDG species. Reductions in C16-and EPA-containing PC and DGTA also correlated with an increase in TAG suggesting that these species are entering TAG biosynthesis, possibly via the dynamic lyso-PC pool (Fig. 8). PDAT with broad substrate specificity have been shown to be a major contributor to TAG biosynthesis in C. reinhardtii under N-deplete conditions [34]. Up-regulation of PDAT and enzymes associated with TAG biosynthesis under N stress have also been reported in P.
tricornutum [35]. However, the pool size of polar lipids is not sufficient to account for the increase in TAG indicating that whilst lipid remodelling occurs under N-deplete conditions, de novo TAG biosynthesis is the major driver behind TAG accumulation. Synthesis of TAG requires a carbon source and the decline in photosynthetic capacity (illustrated here by the reduction in galactolipids) with N depletion, indicates that the source of this carbon must be from the remaining photosynthetic capacity and stored carbohydrates. This observation is supported by the multivariate PCA analysis ( Fig. 6) and has been reported previously by Abida et al. in the Pt1 ecotype of P. tricornutum [28] in contrast to a previous report which proposed betaine lipids as major contributors to TAG formation [31]. An effect that is likely due to differences in growth conditions.

Conclusions
The overall goal of this work was to evaluate the functional role of PtDGAT2B in the synthesis and assembly of TAG. In this study we have successfully demonstrated that PtDGAT2B is an important contributor to TAG biosynthesis and a useful tool to produce high accumulating lipid strains for industrial biotechnology. By overexpressing PtDGATB we have successfully shown how an engineered transgenic strain of P. tricornutum can accumulate elevated TAG levels under N-replete growth conditions. Moreover, the application of a multigene approach, co-expressing DGAT2B with the heterologous Δ5-elongase from the picoalga O. tauri, produced a significant increase in DHA containing TAG, confirming the broad substrate specificity of this DGAT. Furthermore, heterologous expression of transgenes did not impact cell growth rates demonstrating the advantage of this approach to engineer improved lipid content and composition. DGATElo cells represent a promising source of DHA that contain enhanced TAG levels without the requirement for N depletion. Although nitrogen stress is usually used to enhance lipid productivity, this study provides evidence that it is possible to increase TAG content without nutrient stress and engineer TAG to contain highly desirable FAs such as DHA and DPA. To match the level of TAG accumulation under N depletion without effecting the cell growth further knowledge of regulatory mechanisms of lipid production is required.

Strains and growth conditions
Phaeodactylum tricornutum ecotype Pt4 was grown in F/2 media [32], at 20 °C with moderate shaking and constant illumination (60 µmol photons.m -2 .s -1 ). For N stress experiments, replete F/2 media contained enriched phosphate (P) and nitrogen (N) and enriched P with no N in deplete conditions as described by Abida et al. [28]. Cells were grown in enriched F/2 media to the log phase of growth. Cells were then pelleted by centrifugation and used to inoculate 50 ml cultures of replete or depleted media at 3 x 10 6 cells/ml. For each strain n=4-9 depending on the cell type. Cells were counted at time points of 24 h, 48 h and 72 h at which point they were harvested by centrifugation, pellets were washed with 3% ammonium formate, pelleted again, snap frozen in liquid nitrogen and freeze dried.

Plasmid design and cloning
Plasmids were constructed containing either a single PtDGAT2B gene or two-gene cassette with PtDGAT2B and OtElo5 genes. The coding sequences for PtDGAT2A, PtDGAT2B and TpDGAT2 were inserted as KpnI-SacI fragments into position 1 in pPHOS vector under the control of either fcpA or EF2 promoter using a multigene vector described previously [23]. The efficacy of the fcpA promoter in Ndepleted and N-replete conditions was tested and compared to the EF2 promoter (Supplementary   Table S3). The codon-optimised for expression in P. tricornutum coding sequence for O. tauri OtElo5 was inserted as BamHI-XbaI fragment into position 2 of pPHOS to generate double-gene construct DGATElo. Constructs were transformed into P. tricornutum via biolistic according to methods previously described [36]. Positive transformants were selected using the ble marker and confirmed by PCR.

Fatty acid analysis
Total FA content of cells was measured by gas chromatography flame ionisation detection (GC-FID) as described previously [23]. Analysis was carried out using Agilent Chemstation software. Total FAMEs were quantified using 17:0 fatty acid standard (Sigma).

Nile Red staining of non-polar lipids
Non-polar lipids and oil droplets were measured using Nile Red (Sigma-Aldrich) fluorescence staining (excitation at 485-512 nm and emission at 590-610 nm) using a fluorescent microplate reader (Fluostar Omega, BMG Labtech) as described previously [37]. 200 µl culture containing 2.5 x 10 6 cells in the exponential phase of growth was stained with 50 µl Nile Red stock (2.5 µg.ml -1 in 25% DMSO).
Blank measurements (cells incubated without Nile Red stain) were subtracted from test samples and total fluorescence was expressed by dividing fluorescence over that of WT cells.

RNA isolation and qRT-PCR conditions
Total RNA was extracted from 10 8 cells using an RNeasy Plant mini kit (Qiagen, Germany) according to the manufacturer's instructions, with minor modifications. Breakage of the algal cells was aided by freezing in liquid nitrogen prior to the addition of the Qiagen kit RLT lysis buffer. The cell mixture was then sonicated for 1 min followed by an incubation on ice for 1 min, this process was repeated twice.
RNase-free DNase (Promega, USA) was used to remove traces of genomic DNA and the RNA was subsequently purified using RNeasy mini kit columns (Qiagen). The concentration of the extracted RNA was determined using a spectrophotometer at 260 nm and 0.5 µg was reverse transcribed into cDNA using the Transcriptor First Strand cDNA synthesis kit (Roche Applied Science, Germany) according to the manufacturer's instructions. RPS, encoding the ribosomal small subunit 30S, was used as the housekeeping gene since the expression level of RPS has previously been shown to be stable in different conditions [38].
All remaining genes and the corresponding primer pairs used, designed with the Primer3Plus program, are listed in Additional file 13: Table S6. Relative gene expression levels were analysed using the 2 _∆∆C T method using a LightCycler® 96 System (Roche Applied Science). Reactions were started by adding 20 ng cDNA to a mixture containing 1 x SYBR Green PCR Master Mix (Roche Applied Science) and 300 nM of the specific primers in a total volume of 10 µl. The thermal-cycling conditions were as follows: 95°C for 10 min and 40 cycles at 95°C for 10 s, 60°C for 15 s and 15 s at 72 °C.
Melting-curve analysis was performed to check for primer-dimer artifacts. Primer efficiencies were estimated by performing cDNA serial dilutions using the slope calculation supplied with the LightCycler® 96 System Software (Roche Applied Science). For each gene, 3 independent biological replicates and two technical replicates were performed.

Neutral lipid staining and visualisation by confocal microscopy
Non-polar lipids and oil droplets were visualised using Bodipy staining and confocal microscopy according to Govender et al. [39]. Cells were grown in nitrogen replete or deplete conditions and nonpolar lipids and neutral lipids were stained after 24 h, 48 h and 72 h with BODIPY. 0.5 ml cells were pelleted at low speed and re-suspended in artificial sea water (Sigma) containing 1 µg/ml BODIPY (diluted from a stock of 1 mg/ml in DMSO). Cells were incubated for 10 minutes, pelleted by centrifugation and washed in sea water. Cells were then re-suspended in 15 µl sea water prior to confocal microscopy (Zeiss LSM 780).

Quantitative lipid analysis
Total lipids were extracted according to Abida et al., with some modifications [28]. 2 mg freeze dried cells were re-suspended in 4 ml propan-2-ol and heated at 80 °C for 10 minutes. The solvent was transferred to a 50 ml stoppered conical flask. The pellet was then extracted by grinding with a glass homogeniser in 2 ml methanol followed by 8 ml chloroform. Flasks containing solvent extract were flushed with nitrogen gas and stirred for 1 h at room temperature. Samples were filtered through glass wool which was then washed with 3 ml chloroform:methanol (2:1). 5 ml 1% (w/v) NaCl was added to initiate biphase formation and the chloroform phase dried under nitrogen gas. Total lipid extracts were resuspended in chloroform:methanol:ammonium acetate (1:2. The peak area of each lipid was normalized to the internal standard and further normalized to the weight of the initial sample. There is variation in ionization efficiency among acyl glycerol species with different fatty acyl groups, and no response factors for individual species were determined in this study; therefore, the values are not directly proportional to the TAG contents of each species.
However, the approach does allow a realistic comparison of TAG species across samples in this study.

Statistical analysis
Principal component analysis (PCA) was done on the complete dataset of 215 lipids in 138 treatment replicates (24 treatment combinations). Data were log transformed after the addition of a small offset calculated as half the minimum non-zero value. PCA was calculated using the correlation matrix, to give equal importance to each variable. The first two principal components account for 48.69%

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.