Towards the synthetic design of camelina oil enriched in tailored acetyl-triacylglycerols with medium-chain fatty acids

Plants of the oilseed crop camelina were metabolically engineered to produce acetyl-triacylglycerol molecules containing medium-chain fatty acids, resulting in tailored lipids with a designed structure that are potentially useful for biofuel and industrial applications.


Introduction
Seed oils are comprised mostly of triacylglycerols (TAGs), energy-dense molecules that consist of three fatty acids esterified to a glycerol backbone (Fig. 1A). The fatty acid composition of the TAG molecular species determines the physical properties of a particular vegetable oil. TAGs from commercially grown oilseed crops typically contain mainly five fatty carbons can be used for the production of detergents, soaps, lubricants, and biofuels (Dyer et al., 2008). MCFAs are mostly found in tropical plants such as coconut and palm kernel, but temperate plants such as the Cuphea genus (Graham, 1989) and the California bay laurel (Umbellularia californica) produce high levels of MCFAs in their seeds. Even shorter acyl groups are possible. Various Celestraceae species, such as Euonymus alatus, synthesise 3-acetyl-1,2-diacylglycerols (acetyl-TAGs), unusual TAGs with a two-carbon acetate group esterified to the sn-3 position (Kleiman et al., 1967;Sidorov et al., 2014;Tran et al., 2017). The sn-3 acetate group means that acetyl-TAGs possess useful physical properties compared to TAGs with a long-chain fatty acid at that position (hereafter referred to as lcTAGs). For example, acetyl-TAGs possess a lower kinematic viscosity compared to lcTAGs, as well as improved cold-temperature properties, suggesting potential applications as improved biofuels and biolubricants (Durrett et al., 2010;Liu et al., 2015aLiu et al., , 2015b.
The identification of enzymes important for the synthesis and incorporation of unusual fatty acids has allowed their production in transgenic plants through the expression of the encoding genes. For example, the E. alatus diacylglycerol acetyltransferase (EaDAcT) is necessary and sufficient for the synthesis of the acetyl-TAGs that accumulate in the seed endosperm and embryo of that species (Durrett et al., 2010). The seed-specific expression of EaDAcT resulted in acetyl-TAG levels ranging from 47-64 mole percent (mol%) in transgenic Camelina sativa lines. Further, EaDAcT expression combined with the suppression of the enzymes for lcTAG synthesis, e.g. diacylglycerol acyltransferase (DGAT1) and phospholipid diacylglycerol acyltransferase (PDAT1), resulted in acetyl-TAG levels as high as 85 mol% in the best transgenic lines (Liu et al., 2015a(Liu et al., , 2015b. Similarly, MCFAs have been synthesised in a variety of transgenic oilseed crops, through the expression of specialised FatB acyl-ACP thioesterases that cause the release of nascent fatty acids before additional cycles of fatty acid synthesis can extend their carbon chain lengths to C16 or longer (Pollard et al., 1991). In a number of different studies, the seed-specific expression of such specialised thioesterases from plants such as the California bay laurel or different Cuphea species resulted in the accumulation of MCFAs in transgenic canola (Brassica napus) seed (Dehesh et al., 1996;Voelker et al., 1996). Additional MCFA-specific FatB thioesterases were identified in C. viscosissima and C. palustris (Tjellström et al., 2013;Kim et al., 2015b). Subsequent expression of individual or different combinations of these specialised FatB thioesterases in camelina resulted in a range of fatty acid combinations of 8:0 to 16:0 in the transgenic seed oil (Kim et al., 2015b; where, for example, 8:0 refers to a fatty acid with 8 carbon atoms and 0 double bonds). However, accumulation of MCFAs in transgenic seeds expressing these thioesterases was much lower compared to the levels present in the species that naturally synthesise these unusual fatty acids. The co-expression of a MCFA-specific lysophosphatic acid acyltransferase (LPAAT) from coconut was successful in increasing laurate accumulation at the sn-2 position and thus the overall levels of 12:0 (Knutzon et al., 1999). Similarly, expression of other MCFA-specific LPAATs from C. viscosissima and C. pulcherrima caused the accumulation of MCFAs at the sn-2 position in transgenic camelina expressing various Cuphea FatB thioesterases (Kim et al., 2015a).
The oilseed crop Camelina sativa has emerged as a useful platform for the synthesis of unusual lipids through biotechnology approaches. Importantly, a modified floral dip transformation method (Lu and Kang, 2008) enables the relatively easy development of different transgenic lines, allowing optimisation of transgene combinations. Multiple selectable markers permit the stacking of different transgenic traits through crossing or retransformation (Shockey et al., 2015). In addition, field growth of transgenic camelina plants has allowed the large-scale production and subsequent functional testing of different lipids resulting from metabolic engineering, including acetyl-TAGs (Liu et al., 2015a(Liu et al., , 2015b, omega-7 fatty acids (Nguyen et al., 2015), and the nutritionally important omega-3 very long-chain polyunsaturated fatty acids eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) (Ruiz-Lopez et al., 2014;Betancor et al., 2015Betancor et al., , 2016. Here, we describe a synthetic-biology approach to metabolically engineer camelina to produce tailored lipid molecules  with a designed structure. By combining the expression of enzymes resulting in the production of MCFAs with those resulting in high levels of acetyl-TAGs, we demonstrate the synthesis of MCFA-containing acetyl-TAGs that have not been found in nature (Fig. 1). Further, some of the transgenic camelina lines were grown in the field to produce these oils for functional testing of physical properties and they provided information to guide subsequent modifications of the lipid structure. Such iterative approaches will lead to targeted production of lipid molecules designed for specific applications.

Plant transformation
Plasmids expressing EaDAcT alone or in combination with DGAT1 RNAi and/or PDAT1 RNAi (Liu et al., 2015a) were modified by replacement of the DsRed selectable marker with the BAR gene that provides resistance to glufosinate. The gene, along with its nopaline synthase promoter, was amplified from the plasmid pBinG-lyBar1 (Nguyen et al., 2013) using primers with the sequences 5´-GCAGAGCTCGATCTCGGTGAC GGGCAGGACCGGA-3´ and 5´-AGCGATCGGCACGCTGCCGCAAGCACTCAGGGC -3´. The amplified product was digested with SacI and PvuI and ligated into the corresponding sites of the binary vectors containing EaDAcT, a process that simultaneously removed the DsRed gene and promoter cassette. The resulting binary vectors (see Supplementary Fig. S1 at JXB online) were introduced into Agrobacterium tumefaciens GV3101 and transformed into the MCFA-producing camelina lines ChFatB2, ChFatB2+CpFatB2, UcFatB1+CnLPAAT and CpFatB2+UcFatB1 (Kim et al., 2015b) using a floral vacuum-infiltration method (Lu and Kang, 2008). T 1 plants were selected by spraying young seedlings four times with 0.01% (w/v) glufosinate. Lines containing one transgenic locus were identified based on a 3:1 segregation ratio for glufosinate resistance in the T 2 generation. Independent lines with high levels of acetyl-TAGs were further propagated in the greenhouse to generate T 4 seed. Large amounts of T 5 seed for oil property analysis were harvested from plants grown

Lipid analysis
Seed lipids were extracted using a modified hexane-isopropanol extraction method (Li et al., 2006) with tripentadecanoin (Nu Check Prep, Waterville, MN) added as an internal standard. Acetyl-TAGs and lcTAGs were quantified by separating those fractions using TLC, followed by transmethylation and gas chromatography as described previously (Liu et al., 2015a). For electrospray ionisation mass spectrometry (ESI-MS) analysis, neutral lipids were isolated by elution from total seed lipid extracts through a small silica column with 99:1 (v/v) chloroform:methanol. Samples were directly infused into an API4000 Triple Quadrupole mass spectrometer (Applied Biosystems) using methods described previously (Bansal and Durrett, 2016b). The total oil content for different lines was determined gravimetrically on oil extracts from 100 mg of T 4 seed.

Positional analysis of TAGs
The sn-2 fatty acid composition of acetyl-TAGs and lcTAGs in T 4 seed oil were determined by using the lipase from Thermomyces lanuginosus (Sigma-Aldrich), which specifically cleaves the sn-1 and -3 ester bonds of TAGs to generate 2-monoacylglycerols (2-MAGs). Samples of TLCpurified acetyl-TAGs or lcTAGs (1.5 mg) were dissolved in 1.0 ml of diethyl ether, after which 5000 U of lipase in 1 ml of 50 mM Tris buffer (pH 7.2) was added. The reaction was incubated at 37 °C for 30 min with constant shaking. Lipids were removed by extracting twice with 3 ml of diethyl ether and separated on boric acid-impregnated silica gel TLC plates using a chloroform:acetone solvent system (80:10, v/v). Next, 10 µg of tripentadecanoin was added to the bands corresponding to 2-MAGs, which were then scraped and the lipids extracted using 2 ml of toluene. After transmethylation using a base-catalysed reaction, the resulting fatty acid methyl esters were quantified using gas chromatography.

Viscosity analyses
Oil was extracted from batches of 30-50 g of T 5 seed by finely granulating in a coffee grinder followed by Soxhlet extraction with hexane. Acetyl-TAGs and lcTAGs were purified using silica gel chromatography in a 64-mm diameter glass column. Approximately 20 g of oil was applied to 400 g of silica; lipids were eluted with a step gradient that changed the proportions of hexane:diethyl ether from 100:0 to 95:5, to 90:10, to 80:20, and finally to 70:30 (v/v). Fractions (50 ml) were tested for purity using TLC with a 70:30:1 (v/v/v) hexane:diethyl ether:acetic acid solvent system. The fractions containing either pure acetyl-TAGs or lcTAGs were combined and the solvent removed using a rotary evaporator. This extraction and purification process was repeated to obtain enough acetyl-TAGs and lcTAGs from each background to quantify viscosity, typically 25-30 ml. The purity of the fractions was confirmed using ESI-MS. The kinematic viscosity of the purified acetyl-TAGs and lcTAGs was measured using a calibrated Ubbelohde viscometer (Cannon Instruments, State College, PA) at 40 °C according to the ASTM D445 method (ASTM International, 2017).

Synthesis of acetyl-TAGs in MCFA-producing camelina lines
To produce acetyl-TAGs containing MCFAs in planta, we introduced EaDAcT into four different camelina backgrounds, ChFatB2, CpFatB2+UcFatB1, CpFatB2+ChFatB2, and UcFatB1+CnLPAAT, which had been previously engineered to produce MCFAs (Kim et al., 2015b). We also combined the expression of EaDAcT with DGAT1-RNAi and PDAT1-RNAi constructs targeting the assembly of lcTAGs, which have previously been shown to increase the accumulation of acetyl-TAGs (Liu et al., 2015a(Liu et al., , 2015b. We found that the transformation efficiency of the MCFA-producing lines was often low; consequently, we only obtained one or two independent lines per combination of plant genotype and transformation vector. For example, only a total of five EaDAcT-stacked lines were obtained in the ChFatB2 background and just three stacked lines in the CpFatB2+UcFatB1 background ( Supplementary Fig. S2). Slightly more lines were obtained when transforming the CpFatB2+ChFatB2 genotype, particularly with the EaDAcT construct that also targeted both DGAT1 and PDAT1 (Fig. 2). Quantification of acetyl-TAGs in homozygous T 3 seeds revealed a wide range of accumulation levels, with one line producing around 70 mol% acetyl-TAGs whereas three other lines accumulated no or low amounts. Similar bimodal results for acetyl-TAG accumulation have been observed before when expressing EaDAcT in Arabidopsis (Liu et al., 2015a); the reason for this phenomenon remains unknown. In contrast to the other three MCFA backgrounds, multiple transgenic lines were obtained in the UcFatB1+CnLPAAT background with all EaDAcT-expressing vectors (Fig. 2). Transformation with EaDAcT alone resulted in acetyl-TAG levels of around 50 mol% in the T 3 seed of the best transgenic lines, within range of previous observations in a wild-type (WT) background (Liu et al., 2015a). Combining EaDAcT expression with the suppression of PDAT1 did not result in higher levels of acetyl-TAGs. In contrast, co-expression of EaDAcT with DGAT1-RNAi led to higher levels of acetyl-TAGs, with approximately 77 mol% in the best line. The combination of EaDAcT expression with RNAi of both DGAT1 and PDAT1 failed to increase these levels further. These results were consistent with our previous work in a WT background where we observed that the addition of an RNAi construct targeting PDAT1 failed to enhance acetyl-TAG levels beyond those observed with just DGAT1-RNAi (Liu et al., 2015a). One potential explanation is that because both genes have been shown to be important for seed development in Arabidopsis (Zhang et al., 2009), the concomitant strong down-regulation of both DGAT1 and PDAT1 is likely harmful to camelina embryo development.

EaDAcT can synthesise MCFA acetyl-TAGs in camelina
We analysed the composition of neutral lipids extracted from the seeds of homozygous T 3 plants using ESI-MS neutral loss scans to identify acetyl-TAG molecular species. The spectra of MCFA-producing lines expressing EaDAcT contained novel lower molecular-mass acetyl-TAG peaks compared to a WT background line expressing EaDAcT (Fig. 3A). The mass of these peaks corresponded to acetyl-TAG molecules containing MCFA. For example, the ammonium adduct of acetyl-dilaurin (indicated by 26:0 in Fig. 3A) causes the novel peak with m/z of 516.5. The presence of MCFAs in these lower molecularmass acetyl-TAGs was further confirmed using ESI-tandem MS (ESI-MS 2 ), which produced daughter fragments indicative of MCFAs (Fig. 3B). Together, these results demonstrated that EaDAcT was able to acetylate MCFA diacylglycerols (DAGs) such as 1,2-dilaurin-sn-3-glycerol and 1,2-dimyristoyl-sn-3-glycerol to generate acetyl-dilaurin and acetyl-dimyristin in camelina seed. Such activities have been shown in vitro in our previous work (Bansal and Durrett, 2016a). We also quantified the fatty acid composition of the acetyl-TAG and lcTAG fractions of high acetyl-TAG-producing lines in the CpFatB2+ChFatB2 and the UcFatB1+CnLPAAT backgrounds. The distribution of the longer fatty acids typically found in wild-type camelina was generally consistent with previous observations (Liu et al., 2015a): acetyl-TAGs were enriched in more unsaturated fatty acids such as 18:3 and possessed lower amounts of very long-chain fatty acids compared to lcTAGs (Fig. 4). However, in both genotypes, acetyl-TAGs possessed lower amounts of MCFAs compared to the lcTAGs produced in the same lines. For example, in the CpFatB2+ChFatB2 background, 10:0 levels were about fourfold lower in acetyl-TAGs compared to lcTAGs. Likewise, in the UcFatB1+CnLPAAT lines, 12:0 levels in acetyl-TAGs were on average half those in lcTAGs (Fig. 4).

MCFAs are inefficiently incorporated at the sn-2 position of acetyl-TAGs
As the sn-3 position of acetyl-TAG is occupied by an acetate group, the incorporation of MCFAs at the other two positions is particularly important to achieve high levels of these fatty acids in acetyl-TAGs. The regio-specific incorporation of MCFAs into lcTAGs and acetyl-TAGs in high acetyl-TAGproducing transgenic lines was therefore quantified through the use of a stereospecific lipase. Consistent with previous results (Kim et al., 2015b), lcTAGs in lines expressing CnLPAAT contained ~20 mol% laurate at the sn-2 position whereas plants lacking this MCFA-specific LPAAT possessed minimal levels of MCFAs at this position (Fig. 5). Likewise, only acetyl-TAGs from lines expressing CnLPAAT possessed MCFAs at the sn-2 position. However, the levels of laurate at the sn-2 position of acetyl-TAGs were only 5 mol%, considerably lower than those in lcTAGs produced in the same lines (Fig. 5). Taken together, these results emphasise the importance of a MCFA-specific LPAAT to increase accumulation of these unusual fatty acids at the sn-2 position of storage lipids in engineered oilseeds (Knutzon et al., 1999;Kim et al., 2015aKim et al., , 2015b. Further, the much lower incorporation of MCFAs in the sn-2 position of acetyl-TAGs relative to lcTAGs also suggests discrimination of EaDAcT relative to endogenous DGAT1 activity for MCFAcontaining DAG substrates. This observation was consistent with earlier work where we demonstrated in vitro that EaDAcT preferentially acetylates DAG molecules containing unsaturated DAGs (Bansal and Durrett, 2016a).

E a D A c T + P D A T 1 -R N A i E a D A c T + D G A T 1 -R N A i + P D A T 1 -R N A i
UcFatB1 + CnLPAAT CpFatB2 + ChFatB2

Transgenic lines possess a lower seed oil content
When determined gravimetrically, the values for seed oil content of the MCFA-producing lines were lower than those of wild-type camelina. For CpFatB2+ChFatB2 this reduction was quite large (73% of WT) whereas only a minimal reduction was observed with UcFatB1+CnLPAAT (Fig. 6). Similar reductions have been noted for other MCFA-producing camelina lines (Iskandarov et al., 2017). The oil content of acetyl-TAG-producing lines derived from these two backgrounds was further reduced by 10-26%, depending on the specific line (Fig. 6). Previous work has shown that acetyl-TAG-producing lines in a WT background also possess slightly lower oil content (Liu et al., 2015a). Interestingly, expression of a MCFA-specific DGAT1 has been shown to rescue the reduced oil content of camelina lines producing MCFAs (Iskandarov et al., 2017).
Here, the development of EaDAcT variants with improved specificity for MCFAs containing DAG might be helpful in overcoming the reduced oil content of these lines. Indeed, there are multiple reports of specialised transferases being able to reverse the reduced oil content caused by the production of unusual fatty acids in transgenic seeds (van Erp et al., 2011;Hu et al., 2012;Li et al., 2012). An alternative strategy could be to overexpress the regulatory factor WRINKLED1 (WRI1), which rescues the reduced oil accumulation in Arabidopsis plants producing hydroxy fatty acids (Adhikari et al., 2016).

Viscosity of MCFA acetyl-TAGs
The viscosity of vegetable oils is an important parameter for their use as fuel, biodegradable lubricant, and other industrial applications. Previously, we demonstrated that acetyl-TAGs possess a lower kinematic viscosity compared to the lcTAGs found in most vegetable oils (Durrett et al., 2010;Liu et al., 2015aLiu et al., , 2015b. Similarly, other studies have shown that TAGs with MCFAs also possess lower viscosities compared to TAGs with longer chains (Valeri and Meirelles, 1997;Geller and Goodrum, 2000). To determine whether acetyl-TAGs enriched with MCFAs would possess lower viscosity than acetyl-TAGs with regular chain lengths, T 4 seeds from two high acetyl-TAG-yielding lines from different MCFA background lines were bulked in the greenhouse and the field to yield enough oil for viscosity testing. The oil was subsequently fractionated using large-scale column chromatography to obtain pure acetyl-TAGs and lcTAGs. The kinematic viscosities (measured at 40 °C according to ASTM D455) of acetyl-TAGs and lcTAGs from wildtype camelina plants were consistent with previous reports (Liu et al., 2015a(Liu et al., , 2015b, with acetyl-TAGs showing a ~30% reduction compared to lcTAGs (Table 1). LcTAGs from UcFatB1+CnLPAAT possessed a slightly lower viscosity of 29.4 mm 2 s -1 than lcTAGs from wild-type plants (30.7 mm 2 s -1 ). However, the viscosity of acetyl-TAGs synthesised in the UcFatB1+CnLPAAT background was essentially the same as that from a wild-type background, with both being 21.9 mm 2 s -1 . Interestingly, lcTAGs from CpFatB2+ChFatB2 possessed a higher viscosity of 34.5 mm 2 s -1 compared to wild-type lcTAGs. Likewise, the kinematic viscosity of acetyl-TAGs from this background was 23.8 mm 2 s -1 , higher than that of acetyl-TAGs synthesised in a wild-type background ( Table 1).
The unchanged or elevated viscosities of acetyl-TAGs containing MCFAs from UcFatB1+CnLPAAT or CpFatB2+ChFatB2 backgrounds, respectively, probably reflects the poor incorporation of MCFAs into these molecules. Initial attempts to enrich the acetyl-TAG fraction for molecules containing MCFAs were unsuccessful. Another, non-exclusive possibility, particularly for the TAGs from the CpFatB2+ChFatB2 background, is that the reduced polyunsaturated fatty acid (PUFA) levels in the acetyl-TAG and lcTAG fractions also contribute to the increase in viscosity. Previous work has shown that double bonds greatly reduce the kinematic viscosity of lcTAGs (Knothe and Steidley, 2005). Likewise, acetyl-TAGs produced in a camelina high-oleic background possess a higher viscosity than acetyl-TAGs containing more PUFAs from a wild-type background (Liu et al., 2015b).

Conclusions
By taking advantage of the combinatorial nature of synthetic biology we have successfully generated tailored lipid molecules not found in nature. However, additional metabolic engineering is needed to improve the incorporation of MCFAs into acetyl-TAGs to further alter the physical properties of these molecules. The recent availability of alternative MCFA-specific LPAAT enzymes (Kim et al., 2015a) offers the possibility to improve MCFA incorporation into the sn-2 position of DAGs and subsequently acetyl-TAGs. However, strategies beyond increasing MCFA content in DAGs will be required given the apparent discrimination of EaDAcT against saturated DAG molecular species (Fig. 4; Bansal and Durrett, 2016a). It might be possible to use directed molecular evolution approaches, similar to those used to identify high-activity DGAT1 enzymes (Siloto et al., 2009), to select for EaDAcT variants with improved specificity for MCFA-containing DAG molecular species. Alternatively, producing unsaturated MCFAs to enable better utilisation by EaDAcT would also result in an increased production of lower molecular-mass TAG species. This latter strategy also has the advantage in that the desired product would be more unsaturated, likely improving the viscosity and cold-temperature properties. Thus, iterative designtest-learn cycles that involve not only increasing the synthesis of tailored lipid molecules in seed oils, but also testing their physical and chemical properties, will lead to the production of lipid molecules designed for specific applications.