High Flux Through the Oxidative Pentose Phosphate Pathway Lowers Efficiency in Developing Camelina Seeds

The low efficiency of maternally-supplied substrate conversion into growth and storage compounds in Camelina sativa embryos is due to high flux through the oxidative pentose phosphate pathway.

Abstract

Many seeds are green during development, and light has been shown to play a role in the efficiency with which maternally supplied substrates are converted into storage compounds. However, the effects of light on the fluxes through central metabolism that determine this efficiency are poorly understood. Here, we used metabolic flux analysis to determine the effects of light on central metabolism in developing embryos of false flax (Camelina sativa). Metabolic efficiency in C. sativa is of interest because, despite its growing importance as a model oilseed and engineering target and its potential as a biofuel crop, its yields are lower than other major oilseed species. Culture conditions under which steady-state growth and composition of developing embryos match those in planta were used to quantify substrate uptake and respiration rates. The carbon conversion efficiency (CCE) was 21% ± 3% in the dark and 42% ± 4% under high light. Under physiological illumination, the CCE (32% ± 2%) was substantially lower than in green and nongreen oilseeds studied previously. ¹³C and ¹⁴C isotopic labeling experiments were used together with computer-aided modeling to map fluxes through central metabolism. Fluxes through the oxidative pentose phosphate pathway (OPPP) were the principal source of CO₂ production and strongly negatively correlated with CCE across light levels. OPPP fluxes were greatly in excess of demand for NAD(P)H for biosynthesis and larger than those measured in other systems. Excess reductant appears to be dissipated via cyanide-insensitive respiration. OPPP enzymes therefore represent a potential target for increasing efficiency and yield in C. sativa.

False flax (Camelina sativa), a member of the Brassicaceae, shows promise for use as a biofuel oilseed crop (Carlsson, 2009; Iskandarov et al., 2014) and is becoming widely used as a model oilseed plant (Lu et al., 2011; Bansal and Durrett, 2016; Malik et al., 2018). The properties that make C. sativa attractive for study as a model oilseed plant include its being a near relative of oilseed rape (Brassica napus, a major plant oil crop), ease of genetic transformation (Liu et al., 2012; Wang et al., 2012; Liang et al., 2013), its close genetic similarity to the model plant Arabidopsis (Arabidopsis thaliana; Liang et al., 2013), having a sequenced genome (Kagale et al., 2014), and its short life cycle and small stature. Although embryos of C. sativa are larger than Arabidopsis and are therefore more amenable to biochemical analyses, they are small enough that nutrient and/or oxygen gradients are likely to be small compared with larger embryos (Borisjuk et al., 2004), especially in culture. As a crop its advantages include a short life cycle that allows it to be grown between winter and summer crops (Allen et al., 2014; Chaturvedi et al., 2018), resilience to water and nutrient stresses (Iskandarov et al., 2014), the composition of its seed oil (Moser, 2010; Berti et al., 2016), and a relatively low susceptibility to disease and insect pests in its growth range (Vollmann and Eynck, 2015).

Despite its advantages, typical agricultural yields for C. sativa are approximately half those of B. napus, and it is therefore grown much less (USDA NASS, 2012). When exposed to light, most oil producing green seeds, including B. napus and soybean (Glycine max), can convert maternally supplied sugars and amino acids into storage products (starch, oil, and protein) with a higher carbon conversion efficiency (CCE) than nongreen seeds (Sriram et al., 2004; Alonso et al., 2007a, 2009b; Lonien and Schwender, 2009; Alonso et al., 2010, Chen and Shachar-Hill, 2012). Here we report that C. sativa is an exception to this trend and its embryos under physiological light have a lower CCE than sunflower (Helianthus annuus) and maize (Zea mays). Light plays an important role in green seed metabolism by improving the efficiency of carbon storage and generating ATP and/or reductant to the developing embryo (Goffman et al., 2005; Allen et al., 2009b). Developing B. napus embryos have a CCE of close to 80% under physiological light and as light becomes more available have increased CCE, biomass, and yield (Goffman et al., 2005). Increased light has been found to lead to increased labeling in fatty acids from ³H₂O in B. napus and soybean embryos, as well as a higher activation of Calvin Benson cycle enzymes (Ruuska et al., 2004) and higher carbon dioxide fixation rates by developing soybean embryos (Allen et al., 2009b). Soybean appears to be optimized for physiological light conditions as its embryos grew similarly at moderate and high light, but gained less biomass under low light conditions (Allen et al., 2009b). An analysis of the metabolic basis for CCE and the capacity to use light during seed development is thus of interest for understanding biosynthesis in seeds and may be useful in crop improvement.

Because CCE results from the sum of all catabolic and anabolic metabolic processes, a quantitative analysis of the fluxes through central metabolism (which conducts the vast majority of biochemical carbon transformation) is needed to decipher its causes. Steady state metabolic flux analysis (MFA) has been used to gain a quantitative understanding of the movement of carbon through central metabolism in plant cells based on ¹³C labeling (Kruger et al., 2012; O’Grady et al., 2012; Shachar-Hill, 2013). Because plant cells are highly compartmentalized, the labeling data must be taken from multiple metabolic products and multiple labeling experiments to accurately define the major central metabolism pathways (Schwender et al., 2004b; Ratcliffe and Shachar-Hill, 2006; Allen et al., 2007, 2009a). One such study found the carbon-to-nitrogen ratio of substrates provided to cultured soybean embryos influenced the activity of metabolic reactions and led to differences in how label was allocated to fatty acids and other products (Allen and Young, 2013). Additionally, a study on Arabidopsis cell cultures showed the fluxes through reactions synthesizing and using reductant was altered by whether assimilated nitrogen was derived from nitrate or ammonium (Masakapalli et al., 2013). These and other MFA studies have revealed the diversity of central metabolism flux patterns that result from which nutritive compounds are available to a cell and point to the potential of embryo metabolism and CCE being shaped by endosperm composition.

Here we used MFA to investigate why C. sativa embryos have low CCE and how their metabolism is affected by light availability. Figure 1 shows an outline of the MFA approach used. First, we developed culturing conditions that mimicked endosperm in planta and resulted in similar embryo growth rates. Next, we labeled developing embryos under zero, physiological, and high light and measured their fatty acid, amino acid, and carbohydrate label contents. For each light level, we implemented four independent substrate labeling strategies to generate flux maps with improved compartmentation and flux resolution than has been possible in other systems. Our findings show that high oxidative pentose phosphate pathway (OPPP) fluxes explain the low CCE and produce excess NADPH, which appears to be dissipated by cyanide-insensitive respiration. Genetically lowering OPPP fluxes and/or alternative oxidase respiration may therefore increase CCE and could provide a means for increasing the yield and productivity of C. sativa.

Figure 1.

Experimental strategy for metabolic flux analysis of developing C. sativa embryos. Culturing conditions were developed to mimic C. sativa in planta growth at physiological light levels, and embryos were cultured with labeled and unlabeled substrates under physiological, dark, and elevated illumination. Metabolite labeling, uptake, and synthesis rates were measured and used in ¹³C-MFA to quantify central metabolism fluxes, which were then computationally analyzed to yield confidence intervals.

RESULTS

Culture Conditions for Embryo Development Were Established, Yielding Steady-State Growth and Composition Matching Those in Planta

To establish culturing conditions for MFA experiments, we determined the composition of the liquid endosperm that surrounds a developing embryo in C. sativa seeds during early-mid maturation (10–16 d after flowering [DAF]; Pollard et al., 2015). ¹H NMR and gas chromatography–mass spectrometry (GC-MS) measurements revealed Glc was the predominant sugar at 138 mM, with lower levels of Suc (12 mm) also present. Ala and Gln were the most abundant amino acids, each present at ∼2 mm (Supplemental Table S1). Accordingly, we tested different levels and proportions of these four metabolites as the carbon and nitrogen sources in culture media. Plant cells respond to ratios of sugars provided (Gibson, 2005); therefore, the ratio of Glc to Suc used in the growth medium was kept the same as in the endosperm liquid. The carbon-to-nitrogen (C:N) ratio among the metabolic substrates to which embryos are exposed also affects their composition and metabolism (Allen and Young, 2013), and this was adjusted in the culture medium to achieve the same protein levels as in embryos growing in planta. The addition of osmoticum (20% polyethylene glycol [PEG]) was found necessary in culture to limit the growth rate to that in planta and to prevent premature germination, as observed in B. napus (Schwender and Ohlrogge, 2002). The full culture medium composition is listed in Supplemental Table S2. The level of light reaching embryos in planta was determined by measuring the absorbance of the silique and seed coat during development, as reported previously for B. napus and soybean (Ruuska et al., 2004; Allen et al., 2009b), and illumination was set accordingly as 10 µmol m⁻² s⁻¹ of green-filtered light, which is the estimated light reaching the embryos when the plant is growing at ∼1000 µmol m⁻² s⁻¹ of photosynthetic photon flux density.

Rates of biomass accumulation in growth in culture and in planta were linear with R² values > 0.92 at 47 and 44 µg d⁻¹ per embryo, respectively, and were not significantly different (Supplemental Fig. S1). Because the number of cells in these post mitotic embryos was not increasing, the linear rates of biomass accumulation indicate embryos in culture and in planta were in metabolic steady state over this time period. Embryo compositions (lipid, protein, and carbohydrate contents per dry weight) showed no significant difference (P > 0.05) between embryos grown in culture and in planta (Supplemental Fig. S2). Embryos contained 35% lipid, similar to levels reported in the field (USDA NASS, 2012), as well as 27% protein and 8% starch. Together, the similarities in growth rates and biomass compositions indicated the culturing conditions closely mimicked those experienced by embryos in planta.

The CCE of C. Sativa Embryos Was Lower than in Other Studied Species at All Light Levels

To determine the effects of light on metabolism during development, embryos were also cultured in the dark and at elevated light levels (50 µmol m⁻² s⁻¹). Figure 2 shows the composition, relative growth rates, and relative rates of total substrate carbon uptake for embryos in culture under three different light levels (0, 10, and 50 µmol m⁻² s⁻¹). Although growth rates were dramatically increased by light (Fig. 2, top), the substrate uptake rates were less affected and actually decreased at higher illumination (Fig. 2, bottom). Light also increases the growth of B. napus embryos (Goffman et al., 2005), although in soybean, increasing light levels beyond those estimated for in planta development does not increase the growth rate (Allen et al., 2009b).

Figure 2.

Growth, composition, substrate uptake, and carbon use efficiency at different light levels. C. sativa embryos were grown in culture during midmaturation (starting at 10 DAF) under zero (0 µmol m⁻² s⁻¹ for 9 d), physiological (10 µmol m⁻² s⁻¹ for 6 d), or high light (50 µmol m⁻² s⁻¹, for 6 d). Top, growth rates are shown as the diameter of the pie charts, which are divided according to the biomass composition: fatty acids (light gray) measured by extracting total lipids, transesterification, and GC-FID analysis; protein (white) extracted and measured using a colorimetric Bradford assay; starch (dark gray) measured after lipid and protein extraction by treatment with α-amyloglucosydase and recorded as Glc released; and the remaining biomass, which was designated as cell wall (black). Bottom, total substrate uptake rates are shown as the diameters of the pie charts and were determined by measuring substrate contents in unused media or media after embryo culture using ¹H-NMR spectroscopy. The proportions of carbon taken up and converted into biomass and released as CO₂ were determined by ¹⁴C measurements after growth with uniformly labeled substrates and are shown in black and white, respectively. Values show average ± sd (n = 12–16).

The rate of uptake of each substrate was measured using NMR spectroscopy to quantify the levels of substrates in the media before and after culture; spectra also indicated there was no significant release of soluble metabolites by cultured embryos and showed the embryos had low CCEs (Supplemental Figs. S3 and S4). These findings were confirmed with experiments in which all the substrates were uniformly labeled with ¹⁴C to the same specific activity, and the ¹⁴C in embryo biomass and in ¹⁴CO₂ released by embryos during culture were measured. The rates of CO₂ emission were determined by comparing the counts in fresh medium, released CO₂, and those incorporated into embryo biomass (Supplemental Fig. S4). In the dark, C. sativa embryos had a CCE of 21% ± 3%, which is much lower than the 60% found for B. napus embryos (Goffman et al., 2005). Other heterotrophically growing seeds and plant cells analyzed to date also have substantially higher CCE values (Chen and Shachar-Hill, 2012). Under physiological conditions (in which light levels, growth rates and biomass composition matched those of embryos grown in planta), C. sativa embryos had a CCE of 32% ± 2%, while the CCE of developing B. napus embryos has been found to be 86% under 50 µmol m⁻² s^-1of light and 83% for soybean embryos under 30 µmol m⁻² s⁻¹ (the level reaching them in planta; Allen et al., 2009b). Under elevated light conditions (50 µmol m⁻² s⁻¹), C. sativa showed a CCE of 42% ± 4%, compared to 94% for B. napus at 150 µmol m⁻² s⁻¹ (Goffman et al., 2005).

Metabolic Fluxes Were Determined using Four Labeling Strategies

To perform steady-state MFA, the tissue must be both in metabolic steady state (as determined above) and in labeling steady state (isotopic steady state). In cultured plant tissue studies, there is usually a small contribution to the end point labeling measurements from unlabeled preexisting biomass (Allen et al., 2009b). The contribution of this was estimated by culturing 10 DAF embryos at 10 µmol m⁻² s⁻¹ for 6 or 8 d and comparing the resultant biomass labeling patterns. Unlabeled biomass made before the labeling period can prevent labeling levels from reaching steady state (Allen et al., 2009b) during the labeling period. This was tested by comparing the average labeling in protein amino acids after 6 d of labeling (55% ± 3.4%) with that after 8 d (58% ± 3.4%). The initial unlabeled biomass was measured as ∼5% of the biomass after culturing, and this was subtracted from the measured ¹³C labeling values. Embryos cultured in the dark, which grew more slowly, were cultured for 9 d instead of 6 to reach steady-state labeling after accounting for preexisting biomass.

Direct measurements of substrate uptake and CO₂ release fluxes were made (Fig. 1) as these are important in distinguishing alternative flux models and in obtaining unique flux maps from fitting the ¹³C labeling data with flux models. ¹³C labeling experiments were conducted at each light level using media in which one of the substrates was replaced with the same concentration of ¹³C labeled substrate. The three light levels (0, 10, 50 µmol m⁻² s⁻¹) were each analyzed using four independent labeling treatments (each in triplicate): (1) 100% U-¹³C Gln, (2) 100% U-¹³C Ala, (3) 100% 1-¹³C Glc, and (4) 80% 1,2-¹³C Glc, 20% U-¹³C Glc. At the end of the culture period, GC-MS and/or NMR were used to measure the ¹³C contents in lipids, proteinogenic amino acids, starch, and cell wall carbohydrate to determine total and positional ¹³C distributions (Supplemental Table S3).

Potential Network Topologies Were Quantitatively Compared

Some metabolites (e.g. His, aromatic amino acids) are synthesized in the plastid from pentose phosphate pathway intermediates (Umbarger, 1978; Maeda and Dudareva, 2012), but there has been uncertainty as to the subcellular location of the oxidative and nonoxidative reactions of this pathway. Previous studies have found evidence that at least some of the reactions are present in both the cytosol and plastid (Kruger and von Schaewen, 2003) and that the reversible nonoxidative reactions foster rapid exchange of pathway intermediates between the cytosol and plastid (Ratcliffe and Shachar-Hill, 2006). Furthermore, Arabidopsis, a close relative of C sativa, has phosphate-translocator proteins capable of transporting pentose phosphates into the plastid (Kruger and von Schaewen, 2003). Consequently, our model placed the nonoxidative reactions in the plastid to account for plastid-specific metabolite synthesis. We tested whether decarboxylation by the OPPP used cytoplasmic or plastidic hexose phosphate as a substrate to synthesize plastidic ribulose 5-phosphate by comparing the goodness of fit of models using one or other of these substrates. The 50 best model fits when the oxidative reactions were in the plastid were all at least 50% better than the best fit when the reactions were in the cytoplast (Supplemental Fig. S5); therefore, we concluded there was insufficient evidence for a significant cytosolic oxidative flux in the pentose phosphate pathway.

Due to its observed importance in green seeds of related plant species B. napus and Arabidopsis (Schwender et al., 2004a; Lonien and Schwender, 2009), we tested if there was evidence for the Rubisco bypass (Schwender et al., 2004a) operating in developing C. sativa embryos by comparing model fits with and without this reaction as an irreversible, free flux in the model. At physiological and high light, permitting carbon flow through the Rubisco bypass improved the overall agreement between modeled and measured data by 5% or less (Supplemental Fig. S5) and predicted much lower fluxes through Rubisco than the OPPP or through Rubisco in B. napus or Arabidopsis (Schwender et al., 2006; Lonien and Schwender, 2009). We concluded there was little or no Rubisco bypass flux in these embryos.

The OPPP Flux Was Strongly Correlated with Inefficiencies in Carbon Use

Most of the substrate carbon was taken up as Glc (Supplemental Fig. S3). MFA revealed nearly all of this carbon was transported to the plastid as hexose phosphate and oxidatively decarboxylated via the OPPP (Fig. 3; Supplemental Table S4). The remaining pentose phosphate reactions led to a substantial resynthesis of hexose phosphate, which was cycled back to OPPP decarboxylation. The cycling of the carbon in this way led to an OPPP decarboxylation flux of over 38,000 nmol C/embryo/d at physiological light levels (Fig. 3; Supplemental Table S4). Because the decarboxylation reaction converts 1 of the 6 hexose phosphate carbons into CO₂, this high flux resulted in 86% of the total CO₂ produced (Fig. 3; Supplemental Table S4). In comparison, the tricarboxylic acid cycle and synthesis of acetyl CoA for fatty acid synthesis only accounted for 13% of the CO₂ emitted. To test whether this unprecedented rate of CO₂ production by oxidative decarboxylation in the OPPP is enzymatically possible, the maximal activity of 6-phosphogluconate dehydrogenase was measured and found to be in large excess over the high fluxes determined by MFA (Fig. 4). We tested whether alternative flux patterns could account for the low CCE by restricting the oxidative flux through the OPPP in the model. To maintain the measured CO₂ emission rates, the resulting model fits had elevated tricarboxylic acid cycle fluxes and were in very poor agreement with the labeling data.

Figure 3.

Flux maps showing carbon fluxes through central metabolism in developing embryos of C. sativa. Embryos were cultured under dark (A), physiological (B), and high light (C) conditions for 6 (A, B) or 9 (C) d with U-¹³C Ala, or U-¹³C Gln, or 1-¹³C Glc, or 80% 1,2-¹³C Glc and 20% U-¹³C Glc (4 labeling experiments per light condition, ≥3 replicates per experiment). The labeling, uptake, and growth rate data were used to quantify the metabolic fluxes by ¹³C-MFA. Arrow sizes correspond to flux intensity and boxes represent metabolites that were modeled as a single pool across compartments. Values are in units of nmol substrate C embryo⁻¹ d⁻¹ and indicate the average net flux ± 90% confidence interval (see “Materials and Methods”). AcCoA, acetyl coenzyme A; AKG, α-ketoglutarate; C, cytosolic; CIT, citrate; CO₂, carbon dioxide; E4P, erythrose 4-phosphate; ext, external; FUM, fumarate; G3P, glyceraldehyde 3-phosphate; HP, hexose phosphate; MAL, malate; OAA, oxaloacetic acid; P, plastidic; P5P, ribose 5-phosphate; PEP, phosphoenolpyruvic acid; PYR, pyruvate; S7P, sedoheptulose 7-phosphate; TAG, triacylglycerol; TP, triose phosphate; Wall, cell wall carbohydrates.

Figure 4.

C. sativa embryos were enzymatically capable of the high OPPP flux. Soluble proteins from C. sativa embryos cultured at 10 and 50 µmol m⁻² s⁻¹ were extracted and assayed for 6-phosphogluconate dehydrogenase (6PGDH) activity by measuring the rate at which NADPH was produced when Glc 6-phosphate was provided as a substrate. The measured activity (dark gray) was compared with the OPPP flux predicted by ¹³C-MFA (light gray). Bars = experimental average ± sd (n = 3; measured), or average net flux ± 90% confidence interval, as determined from the 20 best fits derived from pseudo data sets (modeled).

As expected from the differences in embryo growth rates and CCE as light levels increased, developing embryos had higher net fluxes toward synthesizing carbohydrates, amino acids, and fatty acids, as well as decreased carbon efflux as CO₂ (Fig. 3; Supplemental Table S4). However, when these fluxes are normalized by total carbon uptake, we can see that the embryos used similar proportions of the carbon allocated to biomass to synthesize the different products (Supplemental Table S4). In contrast, on average, the proportions of carbon uptake released as CO₂ varied across the growth conditions from ∼86% in the dark to ∼58% in the light, and the pentose phosphate reactions varied by close to twofold. This supports the hypothesis that OPPP decarboxylation flux greatly influenced the observed CCE. Indeed, linear regression revealed the OPPP flux explained essentially all of the variation in CCE in across the three light levels (Fig. 5).

Figure 5.

High OPPP flux explains the low carbon use efficiency of developing C. sativa embryos. C. sativa developing embryos were cultured under dark, physiological (10 µmol m⁻² s⁻¹), and high light (50 µmol m⁻² s⁻¹) conditions, and fluxes quantified by ¹³C-MFA. CCE was calculated as the percentage of carbon taken up as substrates (sugars and amino acids) that was used to synthesize biomass components. The difference between the total carbon flux through the OPPP decarboxylation reaction (releasing CO₂) and through the Rubisco bypass (refixing CO₂) fluxes was normalized by total carbon influx to allow these fluxes to be comparable across light conditions and species. The correlation between CCE and the OPPP flux for the C. sativa (without Rubisco flux) was analyzed by linear regression, and the values for B. napus developing embryos were obtained using the average carbon fluxes determined from ¹³C-MFA by Schwender et al. (2006; triangle). Squares represent C. sativa averages as determined from the 20 best fits derived from pseudo data sets of the 3 light levels (see “Materials and Methods”). The 90% confidence intervals for these are displayed as error bars within the squares.

DISCUSSION

Central metabolism in plants is more complex than in other eukaryotic taxa because there is greater subcellular compartmentation of metabolites and reactions, and because reactions and entire subnetworks can be duplicated (Kruger and Ratcliffe, 2008; Allen et al., 2009a; Masakapalli et al., 2010). The analysis of fluxes through central metabolism is further complicated by the uncertainty of the contribution of light. This complexity has generally been tackled in MFA studies by the use of more labeling measurements. This involves analyzing more products, routinely including carbohydrates and lipids (Allen et al., 2009a), whereas most studies of prokaryotes and many analyses of nonplant eukaryotes to date have depended on labeling in proteins alone. Further, MFA studies of plants in recent years have made greater use of subcellular compartment-specific analytes, particularly to differentiate cytosolic from plastidic fluxes, including fatty acids (e.g. longer than C18 versus C18 and shorter) and carbohydrates (e.g. starch versus cell wall or protein glycosylation) than has been typical in studies of other eukaryotes (Sriram et al., 2004; Allen et al., 2007). MFA analyses of plant systems have also obtained more labeling data using different substrate labels and label combinations. For example, Masakapalli et al. (2010) used three differently labeled forms of ¹³C Glc in addressing the localization of the OPPP reactions in Arabidopsis cell suspensions. Almost all prokaryote studies have only used one substrate label, although it has been shown that even in these single-compartment systems, the precision and accuracy of flux estimates can be substantially improved by using multiple labeling strategies (Crown et al., 2015). Here, we performed experiments with four substrate labeling strategies, analyzed labeling in compartment specific carbohydrate and lipid reporters as well as in proteins, and we measured substrate uptake rates and output fluxes, including CO₂, and used stoichiometric balance constraints for lipid component synthesis, protein composition, and total nitrogen uptake versus utilization in biosynthesis. Together with optimization of analytical methods to increase precision, this enabled us to (1) distinguish some, although not all, of the key metabolic pools and reactions duplicated between compartments (Fig. 3); (2) test alternative models of reaction localization and structure using the goodness of fit to the total data (Supplemental Fig. S5); and (3) obtain narrow confidence intervals for fluxes, with most of the estimated flux values at all light levels having 90% confidence ranges that were less than 10% of its mean value (Fig. 3; Supplemental Table S4).

The linear growth rates and biomass compositions indicated the cultured embryos were in metabolic steady state and produced storage products at the same rate as in planta (Supplemental Figs. S1 and S2). Substrate uptake and growth rate measurements indicated developing C. sativa seed embryos have strikingly low CCE (Fig. 2), a result confirmed by ¹⁴CO₂ emission and label uptake measurements (Supplemental Fig. S4). Flux analysis showed this inefficiency was due to extremely high flux through the decarboxylation reaction of the OPPP (Fig. 3; Supplemental Table S4), and the capacity of C. sativa embryos developing in planta to conduct this high rate of OPPP was confirmed with enzymatic activity assays (Fig. 4). The OPPP rate for C. sativa embryos is markedly higher, relative to the rates of substrate uptake and growth, than for other green and nongreen seeds (Fig. 6). The proportion of CO₂ released by the major decarboxylation fluxes in central metabolism (tricarboxylic acid cycle, conversion of pyruvate to acetyl CoA for fatty acid synthesis, OPPP, and recapture by Rubisco) varies markedly in developing seed tissues of crop species. The finding of high OPPP fluxes in C. sativa seed embryos is particularly striking because the OPPP rates reported in other oil producing seed tissues are lower in seeds that, like C. sativa, are green during development than in nongreen seeds (Fig. 6; Sriram et al., 2004; Schwender et al., 2006; Allen et al., 2009b; Alonso et al., 2011). Modeling also supports the idea that OPPP flux in developing embryos is mostly or entirely in the plastid because models with the OPPP in the plastid accounted for the labeling patterns substantially better than models with a cytosolic location (Supplemental Fig. S5). However, duplication of OPPP reactions in the cytosol and plastid and some degree of partitioning of OPPP fluxes between them cannot be excluded (Masakapalli et al., 2010).

Figure 6.

Carbon allocation and CO₂ emission by different metabolic processes in developing seed tissues of six crop species. CCE is represented by the proportion of carbon retained as biomass (striped segments), whereas the proportions emitted as CO₂ by the major contributing processes reflect fatty acid synthesis (dark gray), the tricarboxylic acid cycle (TCA; black), the OPPP (light gray). White cutouts represent Rubisco flux that recaptures a portion of the CO₂ emitted. Metabolic fluxes quantified with ¹³C-MFA were analyzed from this and previous studies: H. annuus (A; sunflower; Alonso et al., 2007a), Z. mays (B; maize embryos; Alonso et al., 2011), G. max (C; soybean; Allen et al., 2009b), B. napus (D; Schwender et al., 2006), Arabidopsis (E; Lonien and Schwender, 2009), and C. sativa (F; this study). FAS, fatty acid synthesis.

The high OPPP fluxes in developing C. sativa embryos suggest the OPPP activity is differently regulated than in other plant systems studied to date. Developmental and environmental cues both regulate the activity of the OPPP in plant cells at the gene expression and posttranslational levels (Kruger and von Schaewen, 2003). The enzymatic activity of Glc 6-phosphate dehydrogenase (G6PDH) in particular is controlled in response to the redox status of the plastid by the formation and reduction of a Cys-Cys bond between conserved Cys residues (Wenderoth et al., 1997) and in the cytosol by phosphorylation of a Thr side chain (Dal Santo et al., 2012). We compared the sequences of G6PDH enzymes of Arabidopsis that are active in the cytosol (AtG6PDH5 and AtG6PDH6; Wakao et al., 2008) and plastid (AtG6PDH1, AtG6PDH2 and AtG6PDH3; Wakao and Benning, 2005) with those of the predicted C. sativa G6PDH proteins (NCBI Resource Coordinators, 2018) using sequence alignment to identify residues conserved among them (Supplemental Fig. S6). The predicted C. sativa proteins most similar to the cytosolically active isoforms of Arabidopsis all contained the conserved regulatory Thr residue, and those most similar to the Arabidopsis plastidic G6PDH protein sequences all contained the conserved pair of Cys residues. Other potential explanations for the high enzymatic activity of G6PDH in C. sativa embryos include elevated protein levels or differences in the activity of regulatory proteins and/or small molecule effectors.

At none of the three light levels did the flux analysis give evidence of significant Rubisco flux in C. sativa embryos. We concluded this from the fact that adding Rubisco activity to the model had very little impact on the overall goodness of fit of the model to the data (Supplemental Fig. S5) and because fitting this model to the data assigned little to no flux to the Rubisco reaction. A further test was conducted to determine whether the embryos refixed significant amounts of the ¹⁴CO₂ they released from labeled substrates by either trapping CO₂ released by the embryos throughout the culture period (limiting the availability of emitted ¹⁴CO₂ for refixation) or only at the end. There was no significant difference in the amount of ¹⁴C in biomass whether or when ¹⁴CO₂ was trapped nor was the amount of ¹⁴CO₂ released dependent on when it was trapped (Supplemental Fig. S4). This conclusion contrasts with MFA analyses of B. napus (Schwender et al., 2006; Junker et al., 2007), soybean (Allen et al., 2009b), and Arabidopsis (Lonien and Schwender, 2009), in which it was concluded that Rubisco operating without the complete Calvin Benson cycle makes a significant contribution to intracellular flux, thereby raising CCE by refixing a significant proportion of the CO₂ released by oxidative decarboxylation reactions in these seeds during the synthesis of storage compounds. Sriram et al. (2004) and Iyer et al. (2008) did not include the Rubisco reaction in their flux model of developing soybean seeds due to the absence of significant dilution of ¹³C labeling in biomass compared with the labeled substrates, indicating there was little fixation of external unlabeled CO₂. However, Allen et al. (2009b) observed that when developing soybean embryos were cultured with ¹³C labeled carbohydrate or amino acid substrates, the C6 of Arg, which is synthesized from intracellular CO₂/HCO₃^-, became significantly labeled, showing that the internal CO₂ was labeled. From this, the observation of label incorporation from external ¹⁴CO₂, and the comparison of alternative models, it was concluded Rubisco was active in soybean embryos and was mostly recapturing CO₂ internally rather that fixing external CO₂. Thus it appears C. sativa is an exception among green seeds studied to date in not having a significant Rubisco flux when exposed to physiologically normal levels of light. This is not necessarily a consequence of a significant OPPP flux, because significant OPPP fluxes with simultaneous Rubisco fluxes have been reported in B. napus (Schwender et al., 2006) and Arabidopsis (Lonien and Schwender, 2009), as well as in a recent MFA study of mixotrophically grown cyanobacteria (Ueda et al., 2018).

Light has strong effects on green seed metabolism during development; it results in faster growth and storage product accumulation and in improved CCE (Goffman et al., 2005). Increased light levels have been shown to increase isotopic labeling in fatty acids of B. napus in the presence of ¹⁴CO₂, as well as higher activation of Calvin Benson cycle enzymes (Ruuska et al., 2004). Light also increases the labeling of amino acids and lipids from ¹⁴CO₂ in soybean (Allen et al., 2009b), but because the embryos of seeds (whether green or non-green) are net CO₂ emitters during development and MFA has not previously been performed at different light levels, the mechanism(s) by which increasing light levels increases CCE (Goffman et al., 2005; this study) has not been established. Possible mechanisms include enabling greater flux through Rubisco, increasing other carboxylation fluxes, supplying reductant for biosynthesis, supplying ATP (thereby reducing the need for oxidative phosphorylation), or some combination of these. A graph of the CCE against the OPPP flux showed that variation in OPPP among C. sativa embryos grown at different light levels accounted for essentially all the variation in CCE (Fig. 5). Furthermore, when the CCE for B. napus is compared with the difference between its oxidative decarboxylation rate (OPPP) and CO₂ fixation rate via Rubisco, the results fall on the same line as C. sativa. Thus the effect of light on CCE is accounted for by its effect on OPPP (or when Rubisco is active, the difference between OPPP and Rubisco fluxes).

Because the OPPP in these seeds operates overwhelmingly in its cyclic mode (Fig. 3; Schuster et al., 2000), which consumes hexose and produces NADPH and CO₂, the decrease in OPPP with increasing light levels suggests that light may act by supplying NADPH. This supply would be produced by oxygenic water splitting at PSII and photosynthetic linear electron transport, which also produces ATP (light driven cyclic electron transport produces ATP alone). The total NADPH produced by catabolism and the total consumed by anabolism can be determined from the MFA results, which show that at all light levels the OPPP produced more NADPH (4250–7333 nmol/embryo/d; Supplemental Table S5) than was needed to meet the demands of biosynthesis in C. sativa embryos (1188–2570 nmol/embryo/d; Supplemental Table S6). Supplemental Tables S5–S7 show the amounts of ATP and NAD(P)H produced and consumed by central metabolism and biosynthesis, including maximal hypothetical production rates (Allen et al., 2009b). Maintenance and transmembrane transport costs have been analyzed by Cheung et al. (2013) for Arabidopsis cells, and we estimated values for these as equal to the sum of consumption of ATP and reductant in both metabolism and biosynthesis (Supplemental Table S6). At all light levels, metabolism produced a large excess of reductant, mainly through the OPPP (Supplemental Tables S5–S7). Consequently, although increased light during development reduced (and presumably replaced to some degree) NADPH production by the OPPP, light was apparently not required to meet biosynthetic demands for NADPH.

Concerning the overall redox balance of metabolism, Allen et al. (2009b) estimated whether light was required to produce reductant during seed development in soybean by comparing the average oxidation state of the substrates taken up by the developing seed tissues (sugars and amino acids) with that of the biomass (principally carbohydrates, proteins and lipids) plus the secreted products (CO₂). In that study, the analysis showed light was not needed to meet overall metabolic demands for reductant. By contrast in developing B. napus embryos, the CCE has been found to exceed 90%, whereas triacylglycerol accumulation accounted for >40% of biomass growth (Goffman et al., 2005), indicating the substrates (sugars and amino acids) were considerably more oxidized than the products of metabolism (from the results of Schwender et al., 2006). Therefore, photosynthetic linear electron transport light must have contributed significantly to the biosynthetic demands of embryo development in that system. In C. sativa, such a comparison (Supplemental Table S8) shows that at all light levels, carbon metabolism (overwhelmingly the OPPP) provided considerably more reductant than was needed to balance the redox state of the cell during growth. This is consistent with the MFA-derived accounting for reductant discussed above.

The potential role of light-driven electron transport in meeting ATP demands can also be assessed by using the MFA results to estimate whether central metabolism alone can meet the ATP demands for biosynthesis. We estimate ATP produced by metabolism (mainly by substrate level phosphorylation and oxidative phosphorylation of net NADH synthesis) at the three light levels (Supplemental Tables S5 and S7) was much less than the amounts consumed (Supplemental Table S6) by the known costs of substrate activation and polymerization (such as hexokinase activity and the ∼4.3 ATP per amino acid unit added during peptide synthesis). This finding is in agreement with that of Allen et al. (2009b) who used this approach to deduce that metabolism alone could not meet cellular ATP demands and that light was needed to meet that shortfall in developing soybean embryos. However, in C. sativa embryos the potential exists for the reductant produced by the OPPP to be used to meet ATP requirements instead of light. Indeed in the dark, the tricarboxylic acid cycle flux and substrate level phosphorylation fluxes (3129 nmol ATP/embryo/d; Supplemental Table S5) cannot meet cellular demands for ATP for metabolism, biosynthesis, and maintenance (4755 nmol/embryo/d), and we conclude that at least in the dark some of the reductant produced in the OPPP is not needed for biosynthesis/redox balance and is instead used to drive oxidative phosphorylation (up to 22,000 nmol ATP/embryo/d could be produced; Supplemental Table S5). This would require the use of NADPH from the OPPP either to drive mitochondrial electron transport directly or to produce NADH for this purpose. The synthesis of NADH from NAD⁺ by the oxidation of NADPH to NADP⁺ could be accomplished by a metabolic cycle such as the malate oxaloacetate shuttle, which consumes plastidic NADPH and produces cytosolic NADH in leaves (Taniguchi and Miyake, 2012). The compartmentation and exchange of malate and oxaloacetate were not resolved in the flux maps, and therefore provide no insight as to whether this shuttle may be operational. In the light, the ATP shortfall could be met by light-driven electron transport. In any case, there is an apparent overall excess of reductant produced above the needs of biosynthesis and the estimated ATP consumption rates.

In preliminary experiments we sought to explain the apparent large excess of NADPH produced by the OPPP after all estimated demands were met. First, the possibility exists that ATP turnover is much higher in C. sativa than in other developing embryos studied to date, perhaps through futile cycling (Alonso et al., 2005, 2007b), and that NADPH from the OPPP is used to meet this demand. We estimated ATP turnover in embryos in the dark (where OPPP flux is highest) by in vivo ³¹P NMR saturation transfer (Lundberg et al., 1990). Embryos grew at the same rate under perfusion in 10 mm NMR tubes as in culture, and the rate of ATP turnover estimated from the NMR analysis was similar to rates observed in studies of growing maize root tips or perfused cultured cells (Roberts et al., 1984; Roscher et al., 1998; Roberts, 1990). Because there did not appear to be an unusually high rate of ATP synthesis, we tested whether excess reductant produced by the OPPP was consumed without ATP production by measuring oxygen consumption in the presence or absence of cyanide or azide. Steady-state oxygen consumption rates in the dark in the presence of either of these inhibitors of oxidative respiration were approximately a third of the rates measured without inhibitors. Thus cyanide insensitive respiration was significant in C. sativa embryos and appeared to consume at least some of the reductant produced by the OPPP. A recent study in Chlamydomonas reinhardtii has shown that mitochondrial alternative oxidase AOX1 serves to dissipate excess NADPH generated in the plastid, thereby protecting the cells from oxidative damage at high light (Kaye et al., 2019). We suggest a similar mechanism operates in C. sativa embryos during development, with the OPPP, rather than light producing excess NADPH.

In conclusion, our results show that increasing light reduced the unusually large OPPP flux in developing C. sativa embryos and thereby raised the efficiency with which substrates provided by the plant were converted to storage products. The OPPP flux greatly exceeds biosynthetic and other requirements for both reductant and ATP, and we suggest a considerable proportion of the NADPH produced, which must be oxidized to regenerate NADP+ for metabolism to function, is dissipated via alternative oxidase activity. Reducing flux through the OPPP by transgenic means therefore has the potential to greatly increase CCE in C. sativa seed development and could significantly increase agricultural seed yield.

MATERIALS AND METHODS

Chemicals

Isotopically labeled substrates (¹³C₅ Gln; ¹³C₃ Ala; 1-¹³C₁ Glc; 1,2-¹³C₂ Glc; and ¹³C₆ Glc) ≥99% ¹³C were obtained from Sigma-Aldrich.

Plant Growth

False flax (Camelina sativa var Sunesson) plants were grown in a growth chamber (BioChambers) at 20°C on a 16-h/8-h light/dark cycle with plants receiving 125 µmol m⁻² s⁻¹ of light. Seeds were placed in a mixture of 25% medium coarseness perlite (Sun Gro Horticulture) and 75% Suremix potting soil (MI Grower Products) that had been autoclaved when moist. Once seeds germinated, a one-sixth strength Hoagland solution was added twice per week. During flowering, stems were tagged daily to track silique age and harvested into a 10% (v/v) Clorox bleach solution to surface-sterilize siliques for culturing.

Light Conditions for Cultures

To establish physiological light conditions for embryos in culture, silique walls at 10 DAF were split and arranged in a single layer within a spectrophotometer cuvette to measure the light transmission efficiency. The same was done with the coats of developing seeds. Percent transmittance was measured at wavelengths from 430 to 663 nm. White light from fluorescent bulbs was passed through a green filter as previously described (Goffman et al., 2005; Pollard et al., 2015) to approximate the transmission spectrum of the silique and seed walls. Light levels (photosynthetic photon flux density) during culture were set to 10 µmol m⁻² s⁻¹ total, which was estimated to be the level to which embryos are exposed in planta during daylight in the growing season.

Culturing Conditions

Liquid endosperm was collected from developing seeds of C. sativa between 10 and 16 DAF, and its composition was analyzed via ¹H NMR (as described below for substrate uptake rates) and GC- MS (as described below for amino acids from proteins). Media was made using the most abundant sugars and amino acids, and included 8 mm Gln, 4 mm Ala, 130 mm Glc, 12 mm Suc, and numerous vitamins (see Supplemental Table S2 for a list of reagents). In addition, 20 mm HEPES was added as a pH buffer and 20% PEG 4000 as an osmoticum (Schwender and Ohlrogge, 2002). Embryos were dissected from seeds at 10 DAF and incubated aseptically at 20°C in 1 mL of medium in 30-mm diameter culture wells (5 embryos per well) containing two layers of filter paper (EMD Millipore), so that embryos were in contact with the liquid medium without being submerged, for 6 d under physiological (10 µmol m⁻² s⁻¹) or high (50 µmol m⁻² s⁻¹) light, or for 9 d in the dark. For labeling experiments, embryos were incubated in media in which one of the carbon compounds was labeled with ¹³C, that is, labeled media contained U-¹³C Gln, U-¹³C Ala, 1-¹³C Glc, or 80% 1,2-¹³C Glc and 20% U-¹³C Glc.

Substrate Uptake Rates

After the incubation period, embryos were removed from wells, rinsed with water, and lyophilized. An internal standard of 280 µL of 10 mm methylphosphonic acid was added to the media in each well, the media was transferred to a tube, and the well was washed twice with 2 mL of distilled water, which was combined with the media and lyophilized. After lyophilization, samples were redissolved in 1 mL of water and the PEG was removed by four extractions with 3 mL of chloroform. The aqueous samples were lyophilized and dissolved in heavy water for ¹H NMR analysis of substrate levels. Samples of fresh media were compared with “spent” media from culture wells to determine uptake rates for the sugars and amino acid substrates. CCE was calculated as previously described (Goffman et al., 2005) as the amount of carbon assimilated into biomass (determined by elemental analysis of %C by the Environmental Stable Isotope Laboratory at Duke University and the dry weight increase during culture) as a percentage of the carbon taken up from the media.

¹⁴C Carbon Balance

¹⁴C labeled carbon substrates were added to the growth medium to provide ∼5 µCi total radioactivity per 6-well plate. The specific activity present of each carbon source was adjusted such that each component was equally represented. Five embryos per replicate were cultured under physiological or high light for 6 d in gas-tight bottles that had small vials adhered to the inner wall. Once bottles were sealed with rubber septa, 1 m KOH was added to the vials via a hypodermic needle to trap CO₂ emitted by the embryos. After incubation, the embryos were rinsed three times with distilled water to remove surface radioactivity and then transferred to scintillation vials where they were dissolved with 2 mL of NCS tissue solubilizing fluid (MP Biomedicals) at 45°C until fully dissolved. The total counts of radioactivity present in the fresh medium was compared with counts in the spent media, the dissolved embryos, and the KOH traps by transferring aliquots of each sample type to scintillation vials containing high flash point cocktail safety-solve scintillation fluid (National Diagnostics). Samples were counted on a liquid scintillation counter (Beckman Coulter), averaged over four trials, and corrected for background and quenching.

Determination of Biomass Components

Total lipids were extracted from lyophilized embryos using 2:1 hexane-to-isopropanol kept at 4°C (Hara and Radin, 1978). Embryos were ground using 5-mm tungsten beads in a Retsch bead mill (Retsch GmbH) run at 30 Hz for 3 repetitions of 5 min. The resulting supernatants were pooled, dried under a stream of N₂ at 60°C, and resuspended in 0.5 mL of 3 m HCl in methanol to obtain fatty acid methyl esters. To ensure the samples were completely dissolved, they were briefly flushed with N₂, incubated for 5 min at 60°C, vortexed, and sonicated. Next, the samples were transmethylated for at least 120 min by incubating at 80°C in a dry oven. The reaction was quenched by adding 250 µL of 5% NaHSO₄ in water, followed by 1 mL of hexane. Samples were mixed vigorously and centrifuged for 10 min at 2000 rpm (560g) at room temperature. The top phase was then removed and dried down at room temperature under N₂. Fatty acid methyl esters samples were resuspended in hexane and run on a 6890N/5973 GC-MS (Agilent) with a 30 m × 0.25 mm internal diameter (ID) × 0.27 µm film thickness Agilent VF-23MS column. Stationary phase was 5% diphenyl /95% dimethylpolsiloxane. Quantification was made possible using the methyl ester of heptadecanoic acid (C17:0) as internal standard at a known concentration.

Proteins were extracted by 20 mm Tris-HCl, pH 7.5; 150 mm NaCl; and 1% SDS buffer at 42°C. Total proteins were then quantified by a DC Protein (Bio-Rad Laboratories) colorimetric assay.

Starch fractions were extracted using 0.1 m acetate buffer at 120°C for 1 h. Starch was then hydrolyzed to Glc by α-amyloglucosidase incubated at 50°C for 1 h. d-Glc was quantified using a total starch assay kit (Megazyme International).

Analysis of Labeling

For lipids, 0.78 mg dry weight lipid from total lipid extracts was suspended in 700 µL of d-chloroform for analysis on a Varian Unity-Plus 500 MHz spectrometer equipped with a 5 mm ¹³C-¹H dual-purpose probe. ¹H spectra were obtained using a 90° pulse angle and a relaxation delay sufficient for all detected carbon sources (10 s) to obtain labeling information. The main peaks of interest in this analysis were those from glycerol.

Next, 0.29 mg dry weight lipid was analyzed as butylamide fatty acid derivatives for average labeling by GC-MS as described previously (Allen et al., 2007) except that the derivatization was carried out at 80°C. Derivatized fatty acids were run on a Trace GC ultra/DSQII GC-MS (Thermo Fisher Scientific) with an Agilent VF-23MS (30 m × 0.25 mm ID × 0.27 µm film thickness) column and a ramp from 100°C to 260°C in 23 min. Resulting peaks were corrected for natural abundance, and labeling was calculated for fragments m/z 115 and 128 in the most abundant fatty acids for C. sativa: C16:0, C18:0, C18:1, C18:2, C18:3, C20:0, C20:1, C22:0. Analysis of these fatty acids gives labeling coverage of the plastidic and cytosolic pools of acetyl-CoA for MFA compartmentation.

Then 0.29 mg dry weight of total extracted lipid was also analyzed for glycerol labeling by first transmethylating with 3 m HCl in methanol as described above. Glycerol was separated in the aqueous phase then lyophilized and derivatized with TBDMS and run on a 6890N/5973 GC-MS (Agilent) with a 30 m × 0.25 mm ID × 0.25 µm film thickness Restek RTX-5MS column. Stationary phase was 5% diphenyl /95% dimethylpolsiloxane (Agilent GCMS – DB5) and a temperature gradient of 135°C to 325°C at 5°C/min was used.

Extracted proteins were lyophilized, hydrolyzed to amino acids in 6 m HCl at 120°C for 24 h, and then purified through a cation exchange column of dowex. Derivatization for GC-MS analysis was accomplished using N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide + 1% tert-Butyldimethylchlorosilane (Neves and Vasconcelos, 1987; Dauner and Sauer, 2000), and samples were run on a 6890N/5973 GC-MS (Agilent) using a DB5 column. Single ion monitoring methods were customized for each set of runs to detect only the amino acids and fragments of interest to increase sensitivity as previously described (Dauner and Sauer, 2000). Peaks were integrated and quantified, and natural abundance and labeling amounts were then calculated (Allen et al., 2007).

Starch hydrolyzed to Glc and lyophilized was incubated in 50 μL of a 2.5% (w/v) methoxyamine HCl in pyridine solution for 45 min at 80°C, followed by trimethylsilyl (Sigma) derivatization (Roessner et al., 2001). Labeling was then analyzed on a 6890N/5973 GC-MS (Agilent) using a DB5 column using the ion fragments with m/z 160 (1-2C) and 364 (1-6C). After all extractions were complete, the remaining pellet contained predominantly cell wall material. This portion was partially hydrolyzed with 6 m HCl at 120°C for 24 h to yield monosaccharides, and then derivatives were analyzed via GC-MS as described for starch.

Statistical Analysis

A linear model ANOVA with Type II sum of squares (aov and car packages, R 3.5.3) was used to test experimental effects (e.g. in planta versus cultured measurements) as fixed effects. Post hoc testing was conducted with the Tukey test at a significance level of 0.05 (lsmeans package, R 3.5.3) to determine whether group means were significantly different.

Model Construction and Constraints

The full model structure with constraints and input data for C. sativa embryos grown under physiological light is given in Supplemental Data Set S1. Briefly, the four labeling treatments were integrated into the MFA model by constructing four identical metabolic networks, each fitted to the data for one set of labeling measurements. By constraining the corresponding fluxes for each network to be equal, the model fitting determined fluxes that best fit all four labeling experiments. The 80% 1,2-¹³C Glc and 20% U-¹³C Glc treatment required two parallel Glc influx reactions. These parallel reactions were further constrained to be within 5% of the treatment ratio (i.e. 4:1) to prevent uptake discrimination between the two Glc substrates. A network nitrogen balance was enforced by constraining the predicted total nitrogenous compound production (i.e. the sum of amino acid production fluxes, scaled by their nitrogen contents) to be within 15% of the nitrogen uptake (from Gln and Ala).

Select reactions were also constrained to better reflect biological restrictions or measurements. First, the biologically irreversible reactions, such as most decarboxylase and all kinase reactions, were constrained to have net fluxes greater than or equal to zero, and an exchange flux equal to zero. Second, to accommodate observed glycerolipid structures, the ratio of glycerol 3-phosphate to total (i.e. cytosolic and plastidic) acyl-CoA production fluxes was constrained to be within 30% of the measured glyceryl-to-acyl moieties ratio. Finally, for fluxes that were directly experimentally derived (e.g. substrate influx, amino acid production), the predicted fluxes were constrained to be within 50% of the measured values.

Calculating Network Fluxes

Metabolic fluxes were quantified using the MFA software ¹³C-FLUX (Supplemental Data Set S2, Wiechert et al., 2001; Wiechert et al., 2015). Initial starting points (sets of flux values) for determining best-fit fluxes were randomly generated with MATLAB R2016a to span a region with radius representing 10% of the total measured carbon influx surrounding the algebraic center of feasible space. Subsequent starting points were generated by random sampling of flux values within 10% distance of the best fit values determined in the first set of optimizations. This procedure was repeated to widely explore the feasible space for best-fit sets of flux values.

Starting flux values were optimized with the Donlp2 program, where model fit was quantified by minimizing the total residuum, which is the sum of squared differences between measured and predicted parameters of labeling, substrate uptake rates, biomass composition, and growth rates. Each squared deviation was divided by the sd of the respective measurements to ensure that more precisely determined data weighed more heavily. To reduce random effects in data weighting due to the effects of chance on the standard deviations of modest numbers of measurement replicates (typically n = 3–6) and also the potential for precise but inaccurate measurements, the measured standard deviations were increased before using as model input. The standard deviations of measured fluxes were increased by 10% of the mean, and the standard deviations of labeling data were increased by adding 1% to the measured deviation (mass isotopomers summed to 100% for each metabolite).

At least 100 optimizations per light level were obtained to maximize the likelihood that the best fit flux sets represented the global best fit. The 20 flux sets with the smallest residuum total were inspected to check that they all converged to similar residuum totals and similar flux values. The starting points corresponding to the 20 lowest residua were identified, and for each one, the optimization was repeated using 50 pseudo datasets of measurements that were generated with Monte Carlo sampling using the average and sd of each measurement. The resulting optima were used to obtain the mean and 90% confidence interval for each flux value.

Enzyme Activity Assay and Sequence Alignment

Enzymatic assays were performed using a similar method previously described for maize embryos (Alonso et al., 2010). One hundred embryos per replicate (3 replicates) at 10 DAF were cultured at physiological or high light conditions for 6 d. At the end of the culturing period, three in planta replicate samples were obtained by harvesting embryos that had developed to the same age in planta, and these were analyzed in parallel with the cultured embryo samples. Embryos were rinsed and frozen in liquid nitrogen. Frozen embryos were ground to a fine powder using 5-mm tungsten beads and a Retsch bead mill (Retsch GmbH). A 1-mL extraction buffer (100 mm HEPES, pH 7.5; 6.2 mm MgCl₂; 1 m dithiothreitol), was added and the embryos were ground again. Samples were spun down at 0°C, and the combined supernatants for each sample were chilled on ice and assayed immediately in a 100-mm HEPES, 0.1 m MgCl₂ solution. Extracts were assayed in a spectrophotometer cuvette and absorption at 342 nm was measured 10 min after each addition. 6-Phosphogluconate dehydrogenase activity was determined by the production of NADPH by first adding NADP to eliminate any endogenous 6-phosphogluconate, then adding 6PG.

Amino acid sequences from cytosolic and plastidic isoforms of G6PDH in C. sativa (NCBI Resource Coordinators, 2018) and Arabidopsis (Wakao and Benning, 2005) were aligned with CLUSTALW (Thompson et al., 1994).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Cultured and in planta embryos grow at similar rates.

• Supplemental Figure S2. Developing embryos in culture and in planta had indistinguishable biomass compositions.

• Supplemental Figure S3. Substrate uptake rates by embryos developing under physiological light levels.

• Supplemental Figure S4. ¹⁴C-labeling confirmed Camelina sativa embryos had low carbon use efficiency.

• Supplemental Figure S5. Testing alternative network structures.

• Supplemental Figure S6. Alignment of Glc 6-phosphate dehydrogenase (G6PDH) sequences from Arabidopsis and C. sativa.

• Supplemental Table S1. Metabolite abundance in liquid endosperm.

• Supplemental Table S2. Nutrient composition of culturing media.

• Supplemental Table S3. Isotopomer labeling of amino acids, carbohydrates, and fatty acids from cultured embryos.

• Supplemental Table S4. Forward and reverse fluxes.

• Supplemental Table S5. Reductant and ATP generated by metabolic fluxes.

• Supplemental Table S6. Reductant and ATP required to synthesize embryo biomass products.

• Supplemental Table S7. Reductant and ATP produced by the carbon substrates.

• Supplemental Table S8. Substrate uptake and biomass product redox fluxes.

• Supplemental Data Set S1. Constraints and data used to model developing Camelina sativa embryos at physiological light.

• Supplemental Data Set S2. Input file to model developing Camelina sativa embryos at physiological light with ¹³C-FLUX.