Engineering compositional changes in oilseeds is typically accomplished by introducing new enzymatic step(s) and/or by blocking or enhancing an existing enzymatic step(s) in a seed-specific manner. However, in practice, the amounts of lipid species that accumulate in seeds are often different from what one would predict from enzyme expression levels, and these incongruences may be rooted in an incomplete understanding of the regulation of seed lipid metabolism at the cellular/tissue level. Here we show by mass spectrometry imaging approaches that triacylglycerols and their phospholipid precursors are distributed differently within cotyledons and the hypocotyl/radicle axis in embryos of the oilseed crop Camelina sativa, indicating tissue-specific heterogeneity in triacylglycerol metabolism. Phosphatidylcholines and triacylglycerols enriched in linoleic acid (C18:2) were preferentially localized to the axis tissues, whereas lipid classes enriched in gadoleic acid (C20:1) were preferentially localized to the cotyledons. Manipulation of seed lipid compositions by heterologous over-expression of an acyl–acyl carrier protein thioesterase, or by suppression of fatty acid desaturases and elongases, resulted in new overall seed storage lipid compositions with altered patterns of distribution of phospholipid and triacylglycerol in transgenic embryos. Our results reveal previously unknown differences in acyl lipid distribution in Camelina embryos, and suggest that this spatial heterogeneity may or may not be able to be changed effectively in transgenic seeds depending upon the targeted enzyme(s)/pathway(s). Further, these studies point to the importance of resolving the location of metabolites in addition to their quantities within plant tissues.
Camelina sativa (L.) Crantz (Brassicaceae) is a promising oilseed crop that has been targeted as a future bioenergy feedstock. Due to its lower production cost relative to other oilseed crops, and its geographic suitability, Camelina could be grown as an economically viable, non-food oilseed source for a number of industrial applications (Carlsson, 2009; Collins-Silva et al., 2011). Camelina seeds typically store between 30 and 40% oil by dry weight with a composition that is abundant in unsaturated fatty acids (approximately 90%); in particular oleic (O, 18:1, 14–18%), linoleic (L, 18:2, 14–20%), linolenic (Ln, 18:3, 28–40%) and gadoleic (G, 20:1, 10–17%) acids, with minor amounts of saturated palmitic (P, 16:0, 5–7%) and stearic (S, 18:0, 2–4%) acids (Ciubota-Rosie et al., 2013; Manca et al., 2013; Nguyen et al., 2013). The large amount of linolenic acid in current cultivars results in a low oxidative stability that presents serious drawbacks for potential biodiesel applications (Ciubota-Rosie et al., 2013). However, engineered Camelina lines with altered seed oil compositions are rapidly achievable through an Agrobacterium tumefaciens-mediated, floral-dip transformation method (Lu and Kang, 2008). Camelina is genetically similar to Arabidopsis and therefore it also acts as a good system for translating findings from a model plant to a crop for applied research (Collins-Silva et al., 2011).
Detailed chemical analyses of the lipids in oilseeds provide a chemical framework for their exploitation through metabolic engineering. Although analyses of seed lipid compositions have offered substantial mechanistic insights into the metabolic pathways for seed oil production, the loss of spatial context for these metabolites limits our understanding of how these pathways might be differentially regulated in different seed tissues. Recent applications of imaging technologies such as mass spectrometry imaging (MSI; for reviews see Horn and Chapman, 2012; Lee et al., 2012) and nuclear magnetic resonance (NMR) imaging (Borisjuk et al., 2012) have provided new insights into the compartmentalization of plant lipid metabolism. Mass spectrometry imaging generates and localizes metabolite ions in situ through MS analysis with high mass accuracy and resolution of selected mass analyzers (e.g. Orbitrap; Hu et al., 2005). Through computational software and accurate mass algorithms, images can be reconstructed to show the detailed distributions of selected metabolites within plant tissues. Using matrix-assisted laser desorption/ionization (MALDI)-MSI we demonstrated that both membrane and storage lipid metabolites in the embryos of Gossypium hirsutum (domesticated upland cotton) were distributed in a somewhat surprising heterogeneous manner (Horn et al., 2012). However, there remain few MSI studies to date on plant tissues, and even fewer on analysis of seed tissues (Lee et al., 2012), requiring additional detailed work to understand this chemical heterogeneity and its significance in seed oil metabolism.
To expand the localization of seed lipid metabolites, MALDI-MSI approaches were used to generate high-resolution visual maps of the Camelina seed lipidome which were compared with whole seed profiles obtained by conventional direct-infusion lipidomics (Bartz et al., 2007). The major focus was on the distribution patterns of individual molecular species of triacylglycerols, the major components of the total seed oil, and their metabolic precursors and major membrane lipids, the phosphatidylcholines. A wide range of transgenic Camelina plants were developed with major changes in seed lipid composition, including high palmitic acid (41%), high oleic acid (70%) (Nguyen et al., 2013) and high linoleic acid (54%) lines. These dramatic changes in fatty acid composition illustrate the ease with which Camelina embryos can accommodate changes to the composition of their seed storage reserves, and provides a comparative context in which to interrogate the influence of genetic changes to seed composition on metabolite compartmentalization. This study provides evidence for the importance of using MSI to evaluate seed metabolism in a spatial context, and helps to inform future strategies that will facilitate further alterations in Camelina and other oilseeds.
Nuclear magnetic resonance imaging of Camelina seed oil
Mature Camelina seed comprises an embryo bent axile consisting of a prominent embryonic axis and two folded cotyledons. While the cotyledons of wild-type Camelina seeds produce the majority of the total oil relative to the embryonic axis, non-invasive examination of intact Camelina seeds by 1H-nuclear magnetic resonance (NMR) imaging (Neuberger et al., 2008), also showed substantial oil content in the embryonic axis, especially in the hypocotyl tissues (Figure 1, Movie S1 in Supporting Information). Similar to NMR profiles of other oilseeds, soybean (Neuberger et al., 2008), rapeseed (Neuberger et al., 2009), cotton (Horn et al., 2012) and tobacco (Fuchs et al., 2013), Camelina seeds showed localized regions enriched in total lipid content, suggesting possible differences in the distribution of the metabolic machinery responsible for oil production and storage in the embryo.
Spatial mapping of phosphatidylcholine and heterogeneity of triacylglycerols in wild-type seeds
We used MSI to spatially map the distributions of individual molecular species of triacylglycerols (TAGs), and one of its major metabolic precursors and abundant membrane lipids, phosphatidylcholine (PC). Representative bright-field images of Camelina cryosections showed a thin seed coat surrounding an embryonic axis oriented parallel with the adaxial and abaxial surfaces of the cotyledons (Figure 2a). The MSI of the relative distribution profiles of individual PC species (Figure 2b–j) in wild-type seed tissues showed striking differences in composition between the cotyledon and embryonic axis tissue. The PC molecules formed both [M + H]+ and [M + K]+ ions (see Table S1 for all molecules identified). Two of the most abundant PC molecular species containing one or two 18:2 (L) acyl chains, i.e. PC-34:2 (Figure 2c) and PC-36:4 (Figure 2g), were enriched within the embryonic axis compared with the cotyledons (see Table S2 for relative percentages). On the other hand, species containing 18:3 (Ln) or 20:1 (G) acyl chains, i.e. PC-34:3 (Figure 2b), PC-36:6 (Figure 2e), PC-36:5 (Figure 2f) and PC-38:4 (Figure 2j), were generally more enriched in cotyledons relative to the embryonic axis. Despite the heterogeneity in several PC molecular species throughout both cross-sections and longitudinal sections, the overall relative PC composition (combining and normalizing the relative mole fraction of each PC molecular species in approximately 1000 spots over the entire embryo for several independent cross-sections) agreed well with analysis of total lipid extracts of intact, mature Camelina seeds by direct-infusion lipidomics and by MALDI-MS (Figure 2k, Table S2).
The MSI of TAG molecular species in wild-type Camelina cross-sections (Figure 3) showed dramatic compositional differences in the embryonic axis relative to the cotyledons, similar to PC (Figure 2). The TAG molecules were detected primarily as [M + K]+ ions with minor amounts of [M + Na]+ ions also present (see Table S1 for all molecules identified). Unlike PC, where 20:1 acyl chains make up a relatively minor portion of overall PC composition, TAG molecular species were abundant in 20:1 content similar to Arabidopsis thaliana seeds (O'Neill et al., 2003). Several TAG molecular species with one or more 20:1 acyl chains, i.e. TAG-56:7 (Figure 3a), TAG-56:6 (Figure 3b), TAG-56:5 (Figure 3c), TAG-56:4 (Figure 3d) and TAG-56:3 (Figure 3e), were enriched in the cotyledonary tissues relative to the embryonic axis (see Table S3 for relative percentages). By contrast, the embryonic axis had relatively more 18:2 and/or 18:3 acyl chains, especially with the absence of 20:1-containing species, i.e. TAG-54:8 (Figure 3f), TAG-54:7 (Figure 3g), TAG-54:6 (Figure 3h), TAG-54:5 (Figure 3i) and TAG-52:4 (Figure 3j). Comparisons of the overall TAG molecular composition profiles analyzed by MSI were consistent with conventional MS analysis of total lipid extracts and by MALDI-MS (Figure 3k, Table S3).
Engineering of Camelina with altered seed oil compositions
The poor oxidative stability of natural Camelina seed oil due to its high linolenic (18:3) content currently limits its use for biodiesel and bio-based lubricants (Ciubota-Rosie et al., 2013). To improve the properties of Camelina oil, a number of metabolic strategies were devised to generate oils enriched in palmitic acid (16:0), oleic acid (18:1) and linoleic acid (18:2). High-palmitate lines (16:0 ≈ 40% total fatty acid content, Table 1) were generated by over-expression of a fatty acid thioesterase (FATB) from Cuphea pulcherrima (CpuFATB1) that has a preference for plastidial 16:0-ACP (acyl carrier protein), effectively altering the available CoA pool for incorporation into membrane and storage lipids. High-oleate lines (18:1 ≈ 70%, Table 1) were generated by RNA interference (RNAi)-mediated suppression of FATTY ACID DESATURASE 2 (FAD2), encoding the microsomal ∆12 oleic acid desaturase, and FATTY ACID ELONGASE 1 (FAE1), encoding the primary 3-ketoacyl-CoA synthase for production of very long-chain fatty acids (≥C20). The result of these modifications is the reduction of 18:2 and 20:1 and a corresponding increase in 18:1 in the Camelina seed oil (Hutcheon et al., 2010; Nguyen et al., 2013). High-linoleate lines (18:2 ≈ 55%, Table 1) were generated by RNAi-mediated suppression of FATTY ACID DESATURASE 3 (FAD3) and FAE1, effectively reducing the production of 18:3 and 20:1, respectively, in the seed. These lines were generated using transgenes under the control of strong seed-specific promoters to limit fatty acid alterations to seed. For the high-oleate lines, the FAD2 RNAi transgene was expressed under the control of the soybean oleosin promoter, whereas expression of all other transgenes in these lines was mediated by the soybean glycinin-1 promoter.
Table 1. Fatty acid composition (wt%) of seeds from wild-type and engineered Camelina sativa seeds
RNAi, RNA interference.
Values for each fatty acid are the average from analyses of seeds from three different plants for each line ± SD.
6.6 ± 0.3
2.9 ± 0.2
22.4 ± 4.2
19.4 ± 1.2
32.6 ± 2.2
1.9 ± 0.2
12.5 ± 0.6
41.0 ± 1.5
5.2 ± 0.4
5.0 ± 0.5
27.6 ± 2.7
13.4 ± 1.1
2.4 ± 0.5
2.0 ± 0.3
7.8 ± 0.8
2.9 ± 0.7
66.3 ± 3.1
7.5 ± 1.3
12.3 ± 1.8
0.7 ± 0.2
2.3 ± 0.9
9.4 ± 0.3
5.6 ± 0.3
22.4 ± 0.2
54.1 ± 1.4
4.8 ± 0.5
1.3 ± 0.2
1.8 ± 0.3
Spatial mapping of PC and TAG heterogeneity in high-palmitate seeds
High-palmitate (16:0) CpuFATB1 over-expressing transgenic Camelina seeds were imaged by MALDI-MSI to gain insight into how alteration of the fatty acid supply might affect the composition and distribution of molecular species of both PC (Figure 4) and TAG (Figure 5). A sharp increase in the overall content of PC-34:2 (Figure 4c,i) with a 16:0 and 18:2 acyl chain led to a surprisingly homogenous distribution of this species in contrast to wild-type seeds (compare with Figure 2c). The increase in 16:0 acyl content in PC coincided with an overall reduction of PC species, with two 18:X acyl chains (Figure 4i), i.e. PC-36:5 (Figure 4e), PC-36:4 (Figure 4f) and PC-36:3 (Figure 4g). Interestingly, the distribution patterns of several major PC species found in wild-type tissues was also altered in this high-palmitate seed, i.e. PC-34:3 (Figure 4b), PC-34:1 (Figure 4e), PC-36:4 (Figure 4f) and PC-36:3 (Figure 4g), albeit at lower overall concentrations. High-palmitate seeds also exhibited a notable increase of PC-32:0 distributed throughout the embryo, although it was present at <5% of the total PC (Figure 4h,i). Despite the overall shift in PC composition in high-palmitate seeds relative to wild-type seeds, the composition of overall PC content visualized through MALDI-MSI was consistent with the analysis of total lipid extracts of high-palmitate seeds by conventional direct-infusion MS (Figure 4i, Table S4).
In the high-palmitate lines, as expected, there was an overall shift in TAG composition (Figure 5j) that mirrored the changes in distribution and overall content of palmitate in PC membrane lipids. The major TAG species in the high-palmitate lines showed overall reduced heterogeneity compared with wild-type seeds. The predominant TAG molecular species produced in the high-palmitate lines had two 16:0 acyl chains, i.e. TAG-50:3 (Figure 5b), TAG-50:2 (Figure 5c) and TAG 50:1 (Figure 5d), which were at negligible concentrations in wild-type seeds (Figure 5j). Other species with a single 16:0 chain also were elevated in the high palmitic lines, including TAG-52:4 (Figure 5e), TAG-52:3 (Figure 5f) and TAG-52:2 (Figure 5g). Surprisingly, a fully saturated TAG, i.e. TAG-48:0 (Figure 5a), which is extremely rare in plant tissues, was produced, albeit at relatively low amounts in the high-palmitic lines. The increase in 16:0 content in seeds resulted in an overall reduction in TAGs with polyunsaturated fatty acids and 20:1 chains such as TAG-54:4 (Figure 5h) and TAG-56:4 (Figure 5i), respectively. The overall TAG composition determined by MALDI-MSI was remarkably consistent when compared with total seed lipid extracts by electrospray ionization (ESI)-MS (Figure 5j, Table S5).
Spatial mapping of the heterogeneity of PC and TAG in high-oleate seeds
Mass spectrometry imaging of FAD2/FAE1 RNAi high-oleate seeds (Figure 6) showed a shift in PC composition (Figure 6h) with the major PC molecular species now containing at least one 18:1 acyl chain, i.e. PC-34:1 (Figure 6d), PC-36:3 (Figure 6f) and PC-36:2 (Figure 6g), all with relatively heterogeneous distribution patterns. Unexpectedly, the major PC species throughout the seed, PC-36:2 with two 18:1 acyl chains, showed much higher levels in cotyledons relative to the embryonic axis (Table S6). The overall levels of polyunsaturated fatty acids in the PC pool were reduced, i.e. PC-34:3 (Figure 6b), PC-34:2 (Figure 6c), PC-36:4 (Figure 6e) and PC-36:3 (Figure 6f), and enriched, relatively speaking, in the embryonic axis. These changes coincided with a major reduction of 20:1 content throughout the embryo (Table S6). Like metabolites in other seeds, the PC composition visualized by MALDI-MSI agreed well with quantification in total seed lipid extracts for the high-oleate lines (Figure 6h, Table S6).
An overall shift in TAG composition in the high-oleate seeds (Figure 7h) mirrored the changes in distribution and overall relative amounts of PC membrane lipids. As expected, the predominant TAG molecular species now contained one or more 18:1 acyl chains, i.e. TAG-52:3 (Figure 7b), TAG-52:2 (Figure 7c), TAG-54:5 (Figure 7e), TAG-54:4 (Figure 7f) and TAG-54:3 (Figure 7g). Similar to the distribution of its (likely) metabolic precursor PC-36:2 (Figure 6g), the major TAG species TAG-54:3 with three 18:1 acyl chains exhibited dramatically higher levels in cotyledons relative to the embryonic axis (Table S7). This increase in oleic-acid-containing TAG species coincided with an overall reduction of TAGs with polyunsaturated fatty acids, i.e. TAG-52:4 (Figure 7a) and TAG-54:6 (Figure 7d), and those with a 20:1 acyl chain similar to the situation with PC species (Table S7). Imaging of TAG composition by MALDI-MS agreed well the analysis of total seed lipid extracts by conventional MS (Figure 7h, Table S7).
Spatial mapping of PC and TAG heterogeneity in high-linoleate seeds
Mass spectrometry imaging of FAD3/FAE1 RNAi high-linoleate seeds of the major PC species containing one or more linoleate moieties (PC-34:2 and PC-36:4) showed that they were mostly uniform throughout the embryo (Figure 8), unlike in wild-type seeds where there was a clear distinction between relative proportions in the cotyledons and embryonic axis (Figure 2). Other minor PC species that were reduced in linolenic or oleic acyl chains, i.e. PC-34:1 (Figure 8c), PC-36:5 (Figure 8d), PC-36-3 (Figure 8f) and PC-36:2 (Figure 8g), were mostly uniform throughout the embryonic axis and cotyledons. Relative quantification of PC composition by MALDI-MSI within sections of high-linoleate lines agreed well with the quantities of PC species measured in total seed lipid extracts by conventional MS (Figure 8h, Table S8).
The changes in TAG distribution patterns and composition in high-linoleate seeds (Figure 9) mirrored the changes in distribution and overall relative composition of PC membrane lipids (Figure 8). The major TAG species produced in these seeds contained at least two 18:2 acyl chains, i.e. TAG-52:4 (Figure 9a), TAG-54:6 (Figure 9d) and TAG 54:5 (Figure 9e), and showed relatively uniform distribution patterns. Additional minor TAGs containing at least one 18:2 acyl chain, i.e. TAG-52:3 (Figure 9b), TAG-54:7 (Figure 9c) and TAG 54:4 (Figure 9f), were also more abundant than in wild-type seeds and more uniformly distributed within the embryo. Imaging of TAGs with 20:1 acyl chains, i.e. TAG-56:5 (Figure 9h), showed an overall reduction throughout the seed. Similar to other genotypes, analysis of the overall TAG composition by MALDI-MSI was consistent with the quantification of TAG molecular species in total seed lipid extracts by conventional MS (Figure 9i, Table S9).
Heterogeneity in wild-type seeds suggest tissue-specific metabolism of TAG
Mass spectrometry imaging of Camelina embryo sections revealed markedly heterogeneous distributions of both membrane (PC) and storage lipids (TAG) that normally would be concealed in conventional chemical analysis of extracts from whole seeds. Indeed, before spatially mapping metabolite distributions within cotton (Horn et al., 2012) and Camelina seeds, we had assumed that the distribution of major lipid species would be relatively uniform throughout the cotyledons and the embryonic axis. Examination of the distribution profiles of TAG and one of its metabolic precursors, PC (Figure 10a), in Camelina seeds adds another layer of intricacy to our understanding of seed lipid metabolism, and in particular the biosynthesis of TAGs (Bates et al., 2009; Li-Beisson et al., 2010; Chapman and Ohlrogge, 2012). This striking heterogeneity within wild-type Camelina seeds suggests potential differences in the mechanisms and regulation of the complex lipid biosynthetic pathways that might operate in different parts of the embryo.
In wild-type seeds, 20:1 and 18:2 acyl chains were enriched in either the cotyledons or embryonic axis, respectively, in both PC and TAG molecular species (Figures 2 and 3). Based on the chemical profiles of PC and TAG in Camelina seeds, a majority of these de novo synthesized fatty acids (FA), 16:0-CoA, 18:0-CoA and 18:1-CoA molecules (Chapman and Ohlrogge, 2012) must be subsequently modified before incorporation into TAG. Two primary biochemical routes involved in generating the final seed acyl composition are modification of fatty acids through elongation of 18:1-CoA to 20:1-CoA, using an extraplastidial fatty acid elongase (FAE1; Figure 10b), or by desaturation through extraplastidial fatty acid desaturases (FAD2 and FAD3) to generate 18:2 and 18:3 acyl chains, respectively, while esterified to PC (Figure 10c). Identifying the relative amounts and the location of these lipid metabolites in wild-type seeds and engineered lines with alterations of these enzymatic steps can enhance the interpretation of the relative activities of these pathways in Camelina embryos.
To illustrate the value of MSI for revealing potential tissue-specific metabolic steps, consider the biosynthesis of two major TAG species (Figure 10d) with striking heterogeneity between the cotyledons and embryonic axis. TAG-56:5 and TAG-54:6 both contain at least two 18:2 acyl chains but probably differ at the sn-3 position with TAG-56:5 and TAG-54:6 possessing a 20:1 or 18:2 acyl chain, respectively. The relative abundance of PC-36:4 with two 18:2 acyl chains throughout the seed suggests that both TAGs could have been synthesized from the same PC-36:4-derived acyl backbone. While it is mechanistically possible to synthesize TAG 56:5 with a 20:1 at either the sn-1 or sn-2 position using a different acyl backbone such as in PC-38:3, the amounts of PC-38:3 were relatively low overall. In the case of the synthesis of TAG-56:5 and TAG-54:6, the patterns of their distribution and the PC-36:4 backbone beg the question whether there might be tissue-specific differences in the final acylation step for these TAGs – either though an acyl CoA-dependent (DGAT, diacylglycerol acyltransferase) or an acyl CoA-independent (PDAT, PC:diacylglycerol acyltransferase) pathway depending on whether the molecules are synthesized in the cotyledons or the embryonic axis.
Studies both in vitro and in vivo have established differences in the specificity and selectivity for esterification of different acyl-CoA molecules to a diacylglycerol (DAG) molecule using DGAT (Jako et al., 2001; Lung and Weselake, 2006) or the removal of the sn-2 acyl chain of PC and subsequent esterification to a DAG molecule at the sn-3 position to generate TAG using PDAT (Dahlqvist et al., 2000; Figure 10a). In Camelina cotyledon tissues, the abundance of 20:1 in TAG molecules (Figure 3a–e) and presumed high selectivity of DGAT for 20:1-CoA over other available acyl CoA molecules (i.e. 18:2) as demonstrated in Arabidopsis seeds (Jako et al., 2001) suggests that the primary pathway for synthesizing TAG-56:5 (and other related TAG molecules) would involve an acyl CoA-dependent DGAT-mediated pathway. An acyl CoA-independent PDAT-mediated pathway could probably still participate partially in TAG biosynthesis in cotyledons as its redundancy has been established in Arabidopsis seeds (Zhang et al., 2009). However, the lack of Tri18:2-TAG suggests that PDAT may not be the favored pathway for the synthesis of major TAG species in cotyledon tissues.
By contrast, in the embryonic axis relatively more 18:2 was incorporated into TAG (i.e. TAG-54:6, Figure 10d) than in the cotyledons. Based on metabolite distributions, a high abundance of both PC-36:4 and PC-34:2 was present in the embryonic axis which could supply 18:2 FA moieties for TAG production. Unlike DGAT, which tends to be more selective for 20:1 over 18:2, PDAT has been shown to be more selective for polyunsaturated and other modified FAs (Ståhl et al., 2004) and is more likely to remove a FA from the sn-2 position of PC. Although the 20:1 content was lower in the embryonic axis relative to cotyledons, there was still a substantial amount of 20:1 incorporated into TAG, indicating that FAE1 is active in the embryonic axis, and 20:1 CoA is present for incorporation into TAG. Even if there were similar amounts of 18:2 and 20:1 in the CoA pools, the higher selectivity and specificity of DGAT for 20:1 also supports the suggestion that PDAT is likely to be more involved in the incorporation of 18:2 into TAG 54:6 in the embryonic axis. Although this possibility remains to be verified experimentally, the availability of spatial information for PC and TAG metabolites affords the ability to develop such hypotheses that otherwise would not be apparent when quantifying lipids in seed extracts. While tissue imaging alone cannot resolve all metabolic possibilities, the demonstration of markedly heterogeneous distributions of metabolites provides important insights that should be taken in consideration when designing experiments for evaluating lipid flux and metabolic engineering within intact seeds.
Engineering of Camelina with altered seed oil compositions
The over-expression of CpuFATB1 in Camelina seeds reduced the heterogeneity of PC and TAG through increased 16:0 production (approximately 40% of total FA), scrambling the differential distributions of major PC and TAG metabolites in wild-type seeds to a more uniform distribution (Figure 10e). Altering the supply of FAs exported from plastids by over-expressing a 16:0-ACP- selective thioesterase, i.e. before elongation to 18:0-ACP (ketoacyl-ACP synthase II) and further desaturation to 18:1-ACP (stearoyl-ACP desaturase) in the plastid, effectively reduced the C18 unsaturated FAs available for incorporation into TAGs. However, palmitate comprised about 40% of the total FA content, and based on imaging experiments and lipidomics in extracts, the predominant PC and TAG species were not fully incorporated with 16:0 (Figure 10e). Imaging of PC-32:0 and TAG-48:0 in high-palmitate lines demonstrated that it was possible to incorporate a 16:0 at each sn position of glycerol in transgenics, albeit at low levels. This suggested a potential metabolic bottleneck for further increases in 16:0 content in Camelina embryos. One possible way to further boost 16:0 levels in this mutant could be to use a lysophosphatidic acid acyltransferase (LPAAT) that is enhanced for 16:0 incorporation to increase its proportion at the sn-2 position, since it appears that PC and TAG still have a major amount of 18:2 at the sn-2 position. However, it is possible that mechanisms are in place to edit out the 16:0 from the sn-2 position of PC to retain critical membrane properties including fluidity and stability which are directly related to membrane desaturation and chain length. Imaging of the TAG-50:X species and reduction of TAG-54:X and TAG-56:X species with one or no 16:0 also suggests the 16:0-CoA pool is large enough to outcompete 18:2, 18:3 and 20:1 for DGAT-mediated esterification at the sn-3 position even with its likely lower enzymatic specificity toward 16:0 CoA; so in this case it appears that enzyme specificity and not enzyme localization sets the upper limits for incorporation of palmitic acid into TAG.
The RNAi-mediated suppression of FAD2 and FAE1 produced Camelina seeds with a high oleate content (approximately 70% of total FA; Nguyen et al., 2013). Lipidomics in total seed lipid extracts clearly showed the high-oleate phenotype; however, without MSI one might predict that these compositional changes would be uniform throughout the seed. Visualization of the heterogeneity in high-oleate lines, in particular of PC-36:2 (Figure 10f) and TAG-54:3 (Figure 10f), suggested the incomplete suppression of FAD2, where there appeared to be less PC-OO and TAG-OOO in the embryonic axis relative to cotyledons. Also, PC-PL, PC-LO and PC-LL were relatively more abundant in the embryonic axis relative to cotyledons, again suggesting that FAD2 was more active in the axis relative to cotyledonary tissues. In this case then, compartmentalization appears to set the upper limit for accumulation of oleic acid, which would not be revealed through analysis of total lipid extracts. It is also possible that the oleosin promoter driving expression of the FAD2 RNAi might also have an influence over this heterogeneity such that oleosin-regulated suppression of FAD2 might be less effective in the embryonic axis.
The RNAi-mediated suppression of FAD3 and FAE1 produced Camelina seeds with high linoleate content (approximately 57% of total FA). Based on results with the high-oleate phenotype, one might predict there would also be substantial heterogeneity in these mutants as a result of incomplete suppression. On the contrary, there was reduced heterogeneity in the major species, i.e. PC-36:4 and TAG-54:6 (Figure 10g). While this reduced heterogeneity in the RNAi-FAD3/FAE1 lines might suggest a more efficiently engineered system than the RNAi-FAD2/FAE1 lines in terms of scrambling the endogenous wild-type heterogeneity, the extent of FA modification was actually less for high-linoleate lines (approximately 57% FA) than for high-oleate lines (approximately 70% FA). As in the case of high-linoleate seeds, it appears that a certain proportion of 18:1 is still directed towards elongation through FAE1 (incomplete suppression) or directly incorporated into PC and TAG without further modification. In addition there was a considerable amount of palmitate in PC (PC-34:2) and TAG (TAG 52:4) in these lines, which would not be affected by the suppression of elongation or desaturation, and these palmitate-containing lipids displayed a heterogeneous distribution, where they were both more concentrated in cotyledonary tissues than in embryonic axis tissues. It might be possible to enhance the expression/activity of KASII that drives the conversion of 16:0-ACP to 18:0-ACP and therefore increase the 18:1-ACP/18:1-CoA pools for further desaturation. On the other hand, when engineering a higher linoleic acid content there may always be a proportion of 18:1 on PC or in the acyl-CoA pool that is shuttled into TAG and unavailable for further modification. Collectively, results with high-linoleate lines benefit from combining interpretations from quantitative lipidomics in extracts with the visual lipidomics in embryo cryosections.
Identification through accurate mass measurements
Comprehensive metabolite identification is becoming increasingly complex as the number of structurally unique metabolites that can be detected continues to grow. Despite the structural variations of endogenously produced metabolites, the sheer number of possible metabolites leads to a number of ions that are close in mass or even isobaric, which may result in misidentification by MS-only approaches. The implementation of MS instrumentation with high mass accuracy (±10 p.p.m.) has substantially improved this preliminary identification of many metabolites especially in the m/z range in which TAG and PC ions are detected (m/z 700–1100) reducing much of the redundancy and misidentification that plagues current characterization of much smaller metabolites. In this study, several TAG and PC molecules were identified within this mass tolerance range and several found at <2 p.p.m. by MALDI-MSI (Table S1). The remarkable agreement of TAG and PC compositions between MALDI-MSI analyses of tissue sections compared with conventional MS-based shotgun lipidomics of whole tissues strongly supports the metabolite identification assignments made in MSI, and emphasizes the ability to quantify from a relative perspective proportions of lipid species within a given class in the face of ion suppression.
In summary, visualizing the chemical heterogeneity in Camelina seeds by MALDI-MSI enhances our understanding and evaluation of genetic engineering strategies to modify lipid synthesis pathways in oilseeds. The distribution patterns of PC and TAG molecular species in cells and tissues help to reinforce their precursor–product relationships in situ that must be considered when redesigning lipid metabolism. The generally good agreement in overall lipid profiles between MALDI-MSI in sections and conventional analysis of total seed lipid extracts validates the utility of imaging approaches for evaluating seed lipid composition in both wild-type and engineered backgrounds. These imaging approaches will continue to improve in the number and types of metabolites analyzed, the spatial resolution for mapping tissues and the quantification capacity. These improvements will contribute to refining our understanding of biochemical pathways in plant tissues for more accurate prediction of the outcomes of metabolic engineering.
Generation of transgenic Camelina lines
The C. sativa L. FAD2/FAE1 hairpin high oleate transgenic line was generated as described in Nguyen et al. (2013). The Camelina FAD3/FAE1 hairpin high-linoleate transgenic line was generated by amplification and cloning of a 323 bp fragment of the FAD3 gene from A. thaliana Col-0 cDNA (see Figure S1 for vector maps and additional cloning details) in addition to cloning of the Camelina FAE1 hairpin cassette as described in Nguyen et al. (2013). For preparation of the Camelina high-palmitate transgenic line, the C. pulcherrima FATB1 acyl-ACP thioesterase (CpuFATB1; GenBank accession KC675176) was identified and cloned by PCR as described in Tjellström et al. (2013). Camelina transformation, DsRed selection (Lu and Kang, 2008) and Basta selection of transformants was done as described previously (Nguyen et al., 2013).
Sample preparation for mass spectrometry imaging
Mature, desiccated Camelina seeds (with seed coat intact) were embedded in 10% (w/v) gelatin (Sigma, http://www.sigmaaldrich.com/) using stainless-steel base molds (Thermo Electron, http://www.thermofisher.com/). The molds were frozen at −80°C for at least 24 h for tissue preservation and solidification. Individual seeds equilibrated to −20°C were sectioned within a Leica CM1950 cryostat (http://www.leica-microsystems.com/) with constant chamber and specimen head temperatures (−19°C). Frozen molds were adhered to a cryostat specimen disc using optimum cutting temperature compound (OCT, Tissue-Tek,www.sakura-americas.com). Longitudinal and cross sections of 30–50 μm thicknesses were collected on glass slides and freeze-dried (Labconco model 79480, http://www.labconco.com/) overnight. Prior to MALDI analysis, sections were equilibrated at room temperature (70°F) in desiccators to minimize moisture content.
Sections were coated with chemical matrix, 2,5-dihydroxybenoic acid (DHB, MALDI-MS grade; Sigma), at 20 mg ml−1 in 70:30 (v/v) methanol:ultrapure (MilliQ UFplus) water. Using the overlay application method, DHB was sprayed onto tissue sections using a SunChrom SunCollect MALDI spotter (Sunchrom Wissenschaftliche Geräte GmbH, http://www.sunchrom.de/; Verhaert et al., 2010) until uniform matrix coverage was observed. A typical spraying protocol required 8–12 successive layers of matrix (each using a typical raster pattern) at flow rates of 8–12 μl min−1. The spray nozzle was held at an offset of 40 mm Z offset with typical stage speeds of the X-axis of 740 mm min−1 and the Y-axis of 195 mm min−1.
Mass spectrometry imaging
A linear ion trap-Orbitrap hybrid mass spectrometer equipped with a MALDI source (MALDI LTQ Orbitrap-XL; Thermo Fisher Scientific, http://www.thermofisher.com/) generated and collected raw mass spectra using the Orbitrap analyzer with xcalibur (v.2.1.0) data acquisition software. The nitrogen laser (MNL 100; Lasertechnik Berlin, http://www.ltb-berlin.de/) provided an output at 337 nm with a maximum energy of 80 μJ per pulse and a repetition rate of 60 Hz (Strupat et al., 2009). The tissue imaging application of LTQ Tune Plus (v.2.5.5) was used to set up the image acquisition from the selected tissue area with a raster step size of 25 μm and laser energy at 12 μJ. The Orbitrap analyzer was used for MS data acquisition in positive ion mode with a typical m/z range of 400–1200 with automatic gain control turned on and a nominal resolution of 60 000.
Data processing and imaging analysis
Raw MSI data were processed as described in Horn et al. (2012) using the Java-based application metabolite imager. Briefly, each acquired scan was associated with its appropriate physical coordinates (X- and Y-positions, representing individual spots) using the dimensions of the selected tissue area established during acquisition. The absolute ion counts of individual lipid species were extracted from each spot analyzed within seed tissues using the theoretical masses for common plant lipids at a constant tolerance of ±10 p.p.m. Relative compositions of each lipid class were calculated using the absolute abundances of each individual species detected at every location. Smoothed images were generated using processed data and colored using a red–green linear scale with a maximum value set according to the most abundant ion within a related set of images to show relative abundances (in mol%) within lipid classes.
Lipidomics of total seed extracts
Fatty acid methyl esters (FAMEs) were prepared by direct transesterification of approximately 25 mg of homogenized seeds from wild-type and transgenic lines using trimethylsulfonium hydroxide (TMSH) reagent as described previously (Nguyen et al., 2013). The FA composition of seeds was measure by analysis of FAMEs by gas chromatography with flame ionization detection as described (Cahoon et al., 2006). All solvents were Optima grade from Thermo-Fisher Scientific.
Total lipid extracts (TLE) were prepared from pooled seed samples (10 mg per biological replicate) from each Camelina line using a modified Bligh and Dyer method described in (Horn et al., 2012) with the following modifications. Mature seeds were homogenized using glass beads in plastic bead beater vials using isopropanol (IPA, 70°C) plus 0.01% butylated hydroxytoluene (BHT; analytical grade, Sigma). For quantification of individual molecular species, a prepared standard mix was added to each extraction containing 200 μg Tri15:0-TAG (Nu-Chek Prep, http://www.nu-chekprep.com/) and 20 μg Di14:0-PC (Avanti Polar Lipids, http://avantilipids.com/; all standards analytical grade).
For ESI-MS analysis, lipid extracts diluted in 1:1 (v/v) chloroform:methanol plus 10 mm ammonium acetate were infused at flow rates of 5–10 μl min−1 into an electrospray source (4 kV spray voltage) of a Thermo TSQuantum triple quadrupole mass spectrometer (Thermo Fisher Scientific). Precursor ions of +184 m/z were used to acquire the PC composition with collision energy of 28 V with acyl chains designated based on prevalent fatty acids. Designations of TAG acyl chains (sn-non-specific) were determined using a series of neutral loss scans for abundant fatty acids. Lipid extracts of wild-type seeds were also analyzed by conventional MALDI-MS by overlaying 5 mg ml−1 DHB matrix.
Nuclear magnetic resonance imaging
High-field experiments were conducted on a vertical 17.6 T Bruker Avance II (Bruker GmbH, http://www.bruker.com/) wide-bore system operating at 750 MHz. Technology developed for imaging of small seeds was applied, as described previously (Fuchs et al., 2013). The Micro 2.5 gradient set used (Bruker) had an inner diameter of 40 mm and a maximum gradient strength of 1 T m−1. The experiments were carried out using a 2 mm inner diameter solenoid coil. A standard three-dimensional spin echo sequence was used to acquire images from the lipid distribution of single seeds (field of view of 2 × 2 × 2 mm3; matrix size of 64 × 64 × 64 μm; isotropic resolution of 31 μm). Our in-house software was used for data reconstruction, while the visualization and modeling was performed with amira.
Research on MSI of oilseeds is supported in part by grants from the US Department of Energy, BER Division, DE-FG02-09ER64812 and Cotton Incorporated (agreement no. 08-395) to KDC. The production of high-oleate and high-linoleate Camelina lines was supported by the US Department of Agriculture–Agriculture and Food Research Initiative 2009-05988 to EBC, and the production of high-palmitic Camelina lines was supported as part of the Center for Advanced Biofuels Systems (CABS), an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, and Office of Basic Energy Sciences under award number DE-SC0001295 to EBC. LB would like to acknowledge the German Federal Ministry of Education and Research (DFG, BO-1917/4-1) for financial support. We thank the Hoblitzelle Foundation for the support of MSI facilities, Kerstin Strupat and Mari Prieto Conaway of Thermo-Fisher Scientific for technical support and UNT Developmental Integrative Biology cluster for cryostat access.