ABSTRACT
The inverse correlation between protein and oil production in soybeans is well-documented; however, it has been based primarily on the composition of mature seeds. Though this is the cumulative result of events over the course of soybean seed development, it does not convey information specific to metabolic fluctuations during developmental growth regimes. Maternal nutrient supply via seed coat exudate measurements and metabolite levels within the cotyledon were assessed across development to identify trends in the accumulation of central carbon and nitrogen metabolic intermediates. Active metabolic operation during late seed development was probed through transient labeling with 13C substrates. The results indicated: i) a drop in lipid during seed maturation with a concomitant increase in carbohydrates, ii) a transition from seed filling to maturation phase characterized by quantitatively balanced changes in the carbon use and CO2 release, iii) changes in measured carbon and nitrogen resources supplied maternally over development, iv) 13C metabolites processed through gluconeogenesis towards sustained carbohydrate accumulation as the maternal nutrient supply diminishes, and v) oligosaccharide biosynthetic metabolism during seed coat senescence at maturation. These results highlight temporal engineering targets for altering final biomass composition to increase the value of soybeans and a path to breaking the inverse seed protein and oil correlation.
One-sentence summary Assessment of temporal changes in metabolism during soybean seed development indicated that lipid turnover during maturation contributes carbon for gluconeogenic production of carbohydrates.
INTRODUCTION
The composition of a seed including protein, oil, and carbohydrate levels establishes its commercial value. In soybean [Glycine max (L.) Merr.], storage protein accounts for 35-40% of seed dry weight, with lipids (i.e., oil) accounting for 18-20%, predominantly as triacylglycerol (TAG) (Adams et al., 1983; Collakova et al., 2013; Li et al., 2015). At a value of approximately $40 billion per year, soybean production is second only to corn in contribution from a crop to the US economy; United States Department of Agriculture (USDA), National Agriculture Statistical Services (NASS). Other biomass components, including carbohydrates, are of less market value, and a subset (i.e., raffinose family oligosaccharides (RFOs)) produced late in development cannot be metabolized for energy by monogastric animals. The RFOs include raffinose and stachyose and are considered anti-nutritional components of livestock feed, therefore detracting from seed value. In soybean, breeding efforts that increased protein content have resulted in lower yields (Mello Filho et al., 2004; Singh et al., 2016; Assefa et al., 2018) and the production of less TAG, indicating a tradeoff between protein and both yield and oil in mature seeds. Breaking the inverse correlations to improve the total seed value without compromising yields are unrealized goals of most breeding and biotechnological efforts.
Central carbon metabolic pathways are responsible for the production of storage reserves including lipids, proteins, and carbohydrates in plants. Though the network of central metabolism is highly conserved across species, there is significant diversity in biomass compositions within plant tissues, indicating that flux through the metabolic pathways can vary extensively. For example, the level of lipids in reproductive organs can range from less than 1% in peas and lentils to greater than 70% in pecans and walnuts and up to 88% in mesocarp tissues such as palm (Dyer et al., 2008; Bates and Browse, 2012; Allen et al., 2015). Other organs such as leaves have low levels of lipids (<5%) in the forms of phospho- and galactolipids for membranes and very little storage lipid in the form of TAG (Lin and Oliver, 2008; Chapman et al., 2012). This variation indicates that steps in the metabolic network are pliable with throughput being context-specific across organs, species, and environments (Allen et al., 2015; Allen, 2016). However, resources available to a developing tissue such as a seed are finite, being constrained by the supply and form of exudates from the seed coat of the maternal plant, usually comprised of sugars (i.e., sucrose, glucose, fructose) and amino acids (glutamine, asparagine, alanine) (Rainbird et al., 1984; Fabre and Planchon, 2000; Schwender and Ohlrogge, 2002; Pipolo et al., 2004; Hernández-Sebastià et al., 2005; Allen et al., 2009). Thus, final seed composition, including oil and protein quantities, is a consequence of the availability of received assimilates and flux through enzymatic steps in metabolic pathways (Allen and Young, 2013; Truong et al., 2013). Understanding the differences in metabolic network flux and operation in tissues and species provides a template to engineer seeds or other organs with value-added compositions.
A quantitative description of temporal changes in metabolic network operation requires experimental methods that can probe stages of seed metabolism precisely and dynamically. Metabolite levels of primary intermediates such as amino acids, sugars, and organic acids decline throughout development while the storage components that include RFOs, lipids, and proteins increase (Fait et al., 2006; Collakova et al., 2013; Li et al., 2015). These levels, however, are routinely reported on a “per gram” basis. As the content of storage components that are considered “inactive/inert pools” in developing seeds increase, the primary metabolites, i.e., “active pools”, are diluted as indicatated by the hypothetical description (Figure 1A). Hence reports of metabolite levels must properly account for dilution due to reserve accumulation when comparing trends over development. Metabolite amounts when compared on a “per seed” basis take into account the increase in storage reserves and may portray more accurately the transient changes in accumulation (Figure 1B vs 1C).
One understudied developmental phase of metabolism is seed maturation. The process of desiccation involves more than drying, as indicated by enhanced enzyme activities and gene expression levels (Angelovici et al., 2010), and has important consequences on final reserve composition. However, hypotheses suggested by gene expression and final compositions require validation. Seed maturation represents ~40% of the entire seed developmental progression (Leprince et al., 2016) during which 10-15% of TAGs are turned over (Chia et al., 2005; Baud and Lepiniec, 2009; Baud et al., 2009) and coincidentally, carbohydrates such as RFOs and cell wall polysaccharides (CWPs) continue to accumulate. The metabolic fate of turned over lipid carbon is not clear, but as the supply of exogenous substrates from the maternal plant ceases, sources of carbon are needed to support biosynthetic demands (Baud et al., 2002; Baud and Graham, 2006; Angelovici et al., 2010). Genes involved in fatty acid oxidation and the glyoxylate cycle are expressed at higher levels late in development (Chia et al., 2005; Fait et al., 2006), suggesting that altered tricarboxylic acid (TCA) cycle metabolism, which can vary extensively in seeds (Schwender et al., 2006; Alonso et al., 2007; Allen et al., 2009), might be necessary to meet differing demands (Rolletschek et al., 2003; Rolletschek et al., 2005; Tschiersch et al., 2011) when seed-based photosynthetic contributions decline (Borisjuk et al., 2005; Fait et al., 2006; Angelovici et al., 2010). Whether changes in mitochondrial respiration (Chia et al. 2005) and peroxisomal metabolism (Salon et al., 1988; Raymond et al., 1992; Eastmond et al., 2000; Eastmond and Graham, 2001) could explain repartitioning of carbon for demand late in seed development and support RFO and CWP production (Kuo et al., 1988; Sánchez-Mata et al., 1998; Fait et al., 2006; Collakova et al., 2013; Gawłowska et al., 2017) is unknown.
The result of lipid decreases and production of RFOs during maturation metabolism is a less valuable grain. Experimental results presented here suggest that turned over reserves, including lipids, provide carbon late in seed development necessary to sustain production of RFOs and CWPs. Temporal changes in seed biomass components, the maternal nutrient contribution, and concentrations of central carbon and nitrogen metabolic intermediates were used to study the changing operation of the metabolic network during seed development. Stable isotopes were used to probe cotyledon and seed coat metabolism specific to the maturation phase to describe changes in carbon partitioning late in development. The differences suggest an engineering opportunity to make seeds with value-added composition by paying attention to temporal effects.
Results
Changes in soybean seed biomass composition during maturation decrease seed value
Seeds were harvested according to size based on contemporary developmental stage descriptions (Naeve, 2005; Licht, 2014) (Figure 2A), removed from pods and seed coats, and weighed (fresh weight; FW). Cotyledons were dried to determine dry weight (DW) and moisture content (Figure 2B). R5 seeds were comprised of 81.5 ± 2.5% moisture, with seed desiccation events reducing this amount to 53.4 ± 0.5 % in R7, 45.9 ± 1.6 % in R7.5, and 12.9 ± 0.8 % in R8 (maturity). Further loss of moisture continued in mature seed over time to less than 9% dry weight categorized as R8b and R8c. The net CO2 release from cotyledons was quantified and indicated a peak in CO2 release at R6 followed by a rapid decline (Figure 2C). The measured CO2 spike was consistent with differences in storage reserve production, including significant CO2 generation when flux from pyruvate to acetyl-CoA enables fatty acid biosynthesis. The lipid production between R6 and R7 over a two week period accounts for 64% of the CO2 generated in this time interval based on the accumulation of lipid in seeds (Figure 3).
During maturation of cotyledons there was a small but insignificant drop in total protein accumulatation from 67.6 ± 3.6 mg seed−1 at R7 to 59.7 ± 4 mg seed−1 at maturity (p = 0.11) and a more dramatic significant change in lipid from 40 ± 1.1 mg seed−1 at R7 down to 30.4 ± 0.2 mg seed−1 at maturity (p = 0.004) (Figure 3). Starch production peaked at R6 (4.3 ± 0.2 mg seed−1) and declined sharply before leveling off at 0.55 ± 0.1 mg seed−1. Sucrose accumulation peaked at 2.2 ± 0.2 mg seed−1 during R7 and remained at that level until maturity, while the RFOs raffinose and stachyose accumulated between R6 and R7 to a maximum of 1.6 and 5.9 mg seed−1, respectively. Total free amino acids increased modestly throughout development (to 0.6 mg seed−1 at R6), and the remaining biomass largely attributed to dietary fiber, including CWP, reached 62.3 ± 7 mg seed−1 at R8. All biomass components increased between R5 to R6, indicating that inverse correlations between individual components are not obligatory when sufficient resources were present. Starch turned over between R6 and R7, when significant lipids (67.5%) and RFOs (84.1%) were being synthesized.
Vegetative carbon and nitrogen sources diminish prior to seed maturation in soybeans
Vegetative parts of the plant are the source of sugars and amino acids for developing seeds during much of development (Hsu et al., 1984; Rainbird et al., 1984; Gifford and John, 1985; Egli and Bruening, 2001; Hernández-Sebastià et al., 2005) and impact final seed composition (Allen and Young, 2013); however, the provisions change as seeds mature. Unlike the significant liquid endosperm present in Brassicaseae, soybean seed coat exudate is barely detectable at any given stage in development and is present as a shiny wet surface on cotyledons that amounts to a few microliters and diminishes with development. To recover the maximum amount of seed coat exudate for measurements and minimize extraction from within the testa, the surface contents of the interior of the seed coat were briefly extracted with an isotonic solution of 20 mM ammonium acetate, pH 6.5. The exterior surface of the developing cotyledon was also briefly immersed in the same buffer to capture and quantify the major contents of the exudate. An overall decreasing trend of total exudate metabolites (Figure 4) was observed during seed development. The total metabolite levels in the exudate decreased significantly as seeds approached maturation phase (R7 and R7.5), consistent with the trends in reserve accumulation (Figure 3) where no new storage proteins or lipids are made past R7, thus indicating a metabolic transition prior to maturation.
Nine amino acids were measured at detectable levels in the exudate of all development stages. Alanine and lysine were detected in R5, R5.5, and R6. Methionine, threonine, tryptophan, serine, glycine, cysteine, tyrosine, and valine were not detected and must be generated from seed-based metabolism during development (Figure 4, Supplementary Table S1). The nitrogen rich amino acids asparagine, glutamine, arginine, and histidine were among the highest in content during peak protein accumulation stages (R5-R6), consistent with prior studies that indicated glutamine and asparagine are important sources of nitrogen for filling soybeans at least during early stages of development (Rainbird et al., 1984; Hernández-Sebastià et al., 2005). The accumulation of nitrogen-rich amino acids including aspartate, glutamate, and arginine during the last stage of development (from R7.5 to R8) could hint at the importance of nitrogen provision for amino acid biosynthesis during germination.
Sucrose was also a significant carbon source through R6 before decreasing (Figure 4, Supplementary Table S1), possibly due to raffinose and stachyose production from the seed coat at R7 and R7.5 stages. Prior reports that investigated RFOs in young developing seeds (Gomes et al., 2005; Kosina et al., 2009) suggested that the exudate contains precursors to RFO biosynthesis, i.e., sucrose, myo-inositol, chiro-inositol, and pinitol, but did not consider stages of development beyond R6. Data in the current study indicate RFOs are produced and exuded from the seed coat during the maturation phase, i.e., R7 and R7.5 (Figure 4, Supplementary Table S1).
In planta levels of metabolites in cotyledons change throughout seed development
The pathway intermediates that are characteristic of stage-specific metabolism were investigated through pool size quantification (metabolite quantities) in cotyledons. Measurements were first obtained on a “per mg DW” basis and converted to amounts per (dried?) seed to account for the increasing inert pools (lipids, protein, and carbohydrates) over the course of development and enable comparison between different stages (Supplementary Table S2). A k-means clustering approach was used to compare changes in trends of metabolite pools over development (Figure 5). All measured amino acids except glutamine clustered into groups 1 and 5, consistent with protein accumulation and a demand for storage protein synthesis plateauing between R7 and R7.5 when storage protein accumulation stopped (Figure 3). The steep increase in amino acid content between R7.5 and R8 (more pronounced for cluster 5) along with the concomitant decrease in storage protein (Figure 3) suggested proteolytic activity during maturation. Cluster 2 consisted of only two metabolites, glucose and fructose, which were elevated at R5 and dropped by R5.5. The variation within this cluster past R5.5 was high, likely due to an increase in glucose content past R6 that was not observed for fructose (Supplementary Table S2). The levels of glucose and fructose in R5 may be a consequence of sucrose breakdown, whereas the increase in glucose at R6 and R7 occured when starch was turning over (Figure 3) supported by the presence of the starch degradation product maltose detected in R6 and R7 (Supplementary Table S2). Maltose, in cluster 3, was similar to the organic acids 2-oxoglutaric acid (2OG), malate, succinate, and fumarate involved in the tricarboxylic acid (TCA) cycle and the sugar phosphates 6-phosphogluconate (6PG), ribose 5-phosphate (R5P), and sedoheptulose 7-phosphate (S7P) involved in the oxidative and reductive steps of pentose phosphate pathway metabolism (PPP). Cluster 3 increased to R6 then declined, similar to measured net CO2 release (Figure 2C), suggesting that CO2 released in R6 resulted from TCA cycle and OPPP activity along with fatty acid biosynthesis, which is necessary to produce lipids during the same time frame.
Metabolites in cluster 3 and cluster 4 (Figure 5) overlapped significantly based on the two-dimentional representation of clusters with principal component analysis (PCA). The common metabolites included those from glycolytic/gluconeogenic pathways and the Calvin-Benson Cycle. Non-overlapping metabolites (sucrose, galactinol, stachyose) were associated with RFO accumulation. The decrease in cluster 4 late in development was consistent with RFO production giving way to CWP biosynthesis (Supplementary Table S2).
Lipid turnover supports biosynthesis of carbohydrate reserves late in development
Both the supply of resources from the exudate and the metabolic events as indicated from the pool size comparisons and altered storage reserve profiles changed as seeds developed. Pool sizes are suggestive of changes in metabolism but the differences in pool sizes cannot be strictly attributed to altered biosynthetic or turnover rates. Given that protein and lipid levels decrease nearing maturity while CWPs and undesirable RFOs accumulate (Figure 3), isotope tracers were used to probe the movement of carbon into and out of metabolic pools. After validation of the culturing approach (Supplementary Data 1), labeled 13C3 glycerol was provided to cotyledons for up to 30 minutes to examine metabolism specific to the lipid degradation at the beginning of maturation phase (R7).
R7 seeds incorporated glycerol to produce triose and hexose phosphates over the course of 30 minutes (Figure 6A), consistent with the capacity of operating enzymes gluconeogenically to convert trioses into carbohydrates. Glycerol was chosen in part because entry into metabolism as dihydroxyacetone phosphate (DHAP) would mimic the source of DHAP from the glycerol-3-phosphate backbone that remains after lipolysis late in development. By 30 minutes, labeled carbon originating from 13C3 glycerol was present at measurable levels in DHAP, GAP, PGA, G6P, G1P, and UDPG (Figure 6A), suggesting that gluconeogenesis may occur late in seed development. Labeling results were further confirmed considering the activity of phosphoenolpyruvate carboxykinase (PEPCK), a key enzyme in gluconeogenic metabolism. PEPCK activity was highest at R7 (Figure 6B), consistent with gluconeogenic activity to supply carbon for carbohydrate metabolism at this stage (Figure 6C).
Isotopic labeling of seed coats indicates production of oligosaccharides that are partitioned to the surface of maturing cotyledons
The presence of RFOs in the exudate was unanticipated and may reflect biosynthetic activities in the seed coat. RFOs in seeds are not required for desication tolerance or germination (Dierking and Bilyeu, 2009; Valentine et al., 2017), and metabolism of the seed coat during development has not been previously described; thus the role of these oligosaccharides remains obscure. During development, the seed coat dry weight is reduced with age, indicating that it may be partially remobilized as the last filial tissue that provides reserves to the seed or could be helpful as an osmotic regulator during germination and the imbibition process. To test the contribution of the seed coat, detached seed coats from the R7 stage were labeled with 13C-sucrose, resulting in the production of 13C-raffinose (see methods for culture system set up) (Figure 7A, B). The soybean cultivar ‘Jack’ and a near isogenic ultra-low RFO line (Jack rs2 rs3) that contained natural variations in the two raffinose synthase genes RS2 and RS3 were used instead of Williams 82 due to the availability of the normal/ultra-low RFO near isogenic line pair (see methods for details) (Hagely et al., 2020). As shown in Figure 7C, a significant amount of 13C incorporation into raffinose within the seed coat was observed in the WT line Jack relative to the ultra-low RFO line (Jack rs2 rs3). This result indicated unequivocally that RFOs result in part from seed coat metabolism and may be an important engineering target to favorably alter soybean seed composition.
Discussion
Mature seed composition is the cummulative effect of events throughout development and is a consequence of: a) supply of carbon, nitrogen, and other resources from the maternal plant, and b) seed-based metabolism. In this study we probed the changes in composition over seed development to explain the reduction in oil and protein levels and the accumulation of oligosaccharides late in development. Though inverse protein to oil relationships have been reported regularly (reviewed in Clemente and Cahoon, 2009; Patil et al., 2017), the levels of these two reserves are not at odds during development (Kambhampati et al., 2019). Levels of both lipid and protein in the seed are highest during the initiation of the maturation phase and coincidentally decline at the time when CWPs are the only biomass component being accumulated (Figure 3) when little exogenous carbon is available (Figure 4) to support biosynthetic demands. Thus, the pull for carbon is not exclusive to oil and protein and their turnover is likely a source for production of other reserves.
Seed-based carbon use efficiency indicates the redistribution of reserves to support metabolism late in development
Carbon conversion efficiency or carbon use efficiency (CUE) has been described in seeds with flux analyses to account for the production of CO2 relative to substrates taken up (Schwender et al., 2004; Alonso et al., 2007; Allen et al., 2009; O’Grady et al., 2012). Though the description provides an indication of carbon lost relative to that converted to biomass, the calculation also reflects the composition of the biomass. Seeds that make large amounts of lipid produce more CO2 as a part of fatty acid biosynthesis than seeds that predominantly make starch. Green seeds capitalize on photosynthesis to improve carbon efficiency (Schwender et al., 2004; Allen et al., 2009). Thus the CUE calculation must take into consideration the metabolic context that differs amongst seeds and tissues.
Analogously, the temporal dynamics of seed metabolism are self-evident from the dramatic change in seed appearance with development. Soybean seeds are green at R5 (Figure 2A) and are capable of productively using available sunlight (Ruuska et al., 2004; Borisjuk et al., 2005; Rolletschek et al., 2005; reviewed in Angelovici et al., 2010) due in part to the contribution of Rubisco-based CO2 fixation (Schwender et al., 2004; Allen et al., 2009). CUEs for stages in metabolism were calculated based on differences in composition and CO2 production over developmental stages, resulting in values of 90% and 76% between R5-R6 and R6-R7, respectively (Supplementary Table S3). Prior reports (Allen et al., 2009) that focused exclusively on seed filling indicated reasonable agreement with these early stages. During this time, the spike in CO2 production occurs when lipid production is high and seeds are starting to transition from green to yellow in color (Figure 2C). TCA cycle metabolism and elevated oxidative PPP were also supported during this interval based on related metabolite clustering analysis (Figure 5). Later in development, differences in CO2 production correlated with parallel drops in TCA cycle and OPPP metabolite levels (cluster 3 of Figure 4).
As seeds continue to develop, the CUE calculation is no longer applicable because the supply of exudate from the maternal plant is exhausted. Instead the balance of one reserve turned over should equate with production of other reserves and CO2, to avoid violation of mass conservation. Developmental staging showed that starch levels drop (Figure 2) to supply other needs including RFO and lipid biosynthesis. From R7 to maturation, the balance of carbon turned over as lipid and protein must account for new production of carbohydrates and CO2 generated. The reported changes in storage reserves (Supplementary Table 3) indicated a balance that was 94% closed. The production of some CO2 late in development may suggest a slight decrease in final seed biomass with desiccation; however, the change in seed weight was not statistically significant.
Interestingly, pertaining to RFO production, our results suggested biosynthesis at multiple locations, with carbon supplied from turned over storage reserves in the cotyledon and also as a result of the withering of the seed coat during dessication. 13C12 sucrose incubation (i.e., a precursor to raffinose) with seed coats indicated production of RFOs (Figure 7) and suggests that the reduction in seed coat biomass during maturation may be analogous to a senescence process where the carbohydrate in the seed coat is converted to RFOs at the surface of the cotyledon.
Gluconeogenic activity is temporally synchronized with lipid degradation to supply turned over carbon for carbohydrate biosynthesis
Steady-state metabolic flux analysis using developing soybean seeds previously indicated that gluconeogenic metabolism does not occur when seeds receive adequate supplies of sugars (Allen et al., 2009). However, the composition of seed exudate late in development indicates that seeds do not receive an extensive supply of sugars from the maternal plant at these stages (Figure 4). As shown with 13C3 glycerol labeling experiments (Figure 5A), enzymes involved in shuttling carbon to hexose phosphates can operate effectively late in development so that carbon can be used to produce carbohydrates (RFOs and CWPs). 13C enrichment occured in intermediates of glycolysis (DHAP, PEP), hexose phosphates (G6P, G1P), and the nucleotide sugar UDPG, which are precursors to carbohydrate production (Figure 6). From the balance of biomass components (Supplementary Table 3), the CO2 loss of 3.3 mg (calculated) could come from PEPCK activity (61%; see Supplementary Table 3 for details) which was highest at R7.
The carbon required for PEPCK activity is likely obtained via the glyoxylate cycle, which utilizes acetyl-CoA derived from repeated deacylation of lipids beginning at R7 and continuing throughout the maturation phase by β-oxidation. The key enzymes required for glyoxylate cycle and β-oxidation, isocitrate lyase (ICL), malate synthase (MS), 3-ketoacyl-CoA thiolase (KAT), and the multifunctional enzyme (MFP) of β-oxidation, were previously shown to increase in activity during the maturation phase of embryo development in Brassica napus, characterized by lipid degradation (Chia et al., 2005). We observed that the glyoxylate levels increased over the course of development, peaking at R7 and R7.5 (cluster 4 of Figure 5, Supplementary Table 2). Further, activity of the glyoxylate cycle and gluconeogenesis are supported by prior measurements of transcript abundance over soybean development (Collakova et al., 2013; Li et al., 2015). The differences in carbon movement between the R6 stage of seed filling and inititation of maturation at R7 are stark and are summarized in Figure 8. Carbon received as sucrose via maternal contribution in R6 is channeled into lipid biosynthesis at R6. As the maternal contribution decreases between R6 to R7, starch turnover may contribute carbon to both lipid and oligosaccharide biosynthesis while TCA cycle operation sustains the energy required for oligosaccharide production. As the seeds reach the R7 maturation phase, turnover of lipids is initiated and the carbon from the glycerol backbones of lipids as well as degraded acyl chains is channeled into cell wall polysaccharides via gluconeogenic and β-oxidative pathways, respectively.
Recent efforts to improve seed quality have targeted lipases to reduce lipid breakdown late in development (Kanai et al., 2019) and RFO biosynthetic steps (Valentine et al., 2017; Hagely et al., 2020) to improve seed compositional traits without significant phenotypic consequences to maturation or germination. Carbohydrates as a whole (RFOs and CWPs) constitue ~40% of final seed composition and are an important sink and potentially rob carbon from the production of oil and protein. Hence, future engineering efforts for increased oil and protein, if focused on manipulating key carbon partitioning pathway nodes that consider all three biomass components, can be fruitful in breaking the perceived inverse correlation. Improved channeling of carbon from malate towards lipid using malic enzyme (Allen and Young, 2013) temporally and increasing the sink strength of developing seeds (Rolletschek et al., 2020) by manipulating the hormone status (Quoc Thien et al., 2016; Kambhampati et al., 2017) represent unrealized potential targets for future soybean improvement.
Materials and Methods
Plant growth conditions and tissue collection for in planta and seed coat exudate measurements
Soybean cultivar, Williams 82, was grown under greenhouse conditions as previously described (Kambhampati et al., 2019). Germinating seeds were transferred to one-gallon pots containing Fafard 4M and grown at 25°C-27°C/21°C-23°C day/night temperatures with greater than 35% humidity and sunlight supplemented by approximately 400-1000 Wm−2 to establish a 14 hr day, 10 hr night photoperiod. Plants were watered daily and received Jack’s 15-16-17 (JR Peters) fertilizer three times a week. Developing seeds were grouped based on fresh weight and visual appearance to determine the developmental stage (Figure 1). At the time of harvest, seeds that were used as controls representing in planta conditions were dissected from pods, the seed coat was removed and cotyledons were flash frozen with liquid nitrogen and stored at −80°C until further use. Cotyledons collected from a single plant were treated as a single biological replicate for all stages of development. Tissue that belonged to each replicate was ground individually using machined home-made stainless-steel hammer-crushing pestle and mortar design. Ground and frozen tissue was then lyophilized and aliquoted for individual biomass component and metabolite measurements.
For exudate experiments an isotonic solution of 20 mM ammonium acetate pH 6.8 was placed on the interior side of excised seed coats and pipeted up and down repeatedly for 10 seconds. In addition, the surface of the corresponding cotyledon was rinsed with the same solution to collect any surface contents. The extracts from each stage were dried using a speed vacuum centrifuge and resuspeneded in water and filtered using 0.45 μm cellulose acetate centrifuge filters (costar®, Corning Inc.) prior to metabolite measurements using LC-MS/MS.
Moisture content, fresh weight, dry weight, and net CO2 measurements
Moisture content was determined using fresh weight and dry weight measurements. Seeds were weighted immediately after harvesting to obtain fresh weights, sliced and dried in an oven. Dried seeds were measured at least three times over the course of several weeks to ensure no moisture remained and the weights plateauted. CO2 measurements were taken from whole soybean cotyledons, after excising the pod walls and seed coats, using a LI-COR 6400 with an attached insect respiration chamber (#6400-89) following manufacturers protocol. Five replicates with three cotyledons each and ten measurements were used with readings taken every 20 to 30 seconds. 30 μE of light was maintained throughout the measurement period to simulate light received by the cotyledons within the pods (Allen et al., 2009).
Normal seed RFO soybean cultivar ‘Jack’ and the ultra-low RFO derivative ‘Jack rs2 rs3’
An ultra-low RFO version of soybean cultivar ‘Jack’ (Jack rs2 rs3) was developed by backcrossing variant alleles of the raffinose synthase 2 (rs2) and raffinose synthase 3 (rs3) genes into Jack (Hagely et al., 2020). The ultra-low RFO phenotype has been defined as raffinose and stachyose content less than 0.70% of seed dry weight (Hagely et al., 2013; Schillinger JA, 2013, 2018).
13C labeled culture system set up and conditions
For culturing system development, cotyledons from specific stages were excised from seed coats under sterile conditions and immediately placed flat face down for each cotyledonary half into 300 μL of sterile culture medium in a 24-well plate. A modified Linsmaier and Skoog medium (Thompson et al., 1977; Hsu and Obendorf, 1982) with Gamborg’s vitamins (Sigma) and 5 mM MES buffer adjusted to pH 5.8 was used as the culture medium and contained 200 mM U-13C6 glucose as the exclusive carbon source. Culturing was performed under 30 – 35 μE continuous light at 26°C, consistent with Allen et al. (2009) and tissue was collected at 5, 10, and 30 mins, in triplicates, for all time course studies described. Untreated samples were also taken in triplicates for 0 timepoint (in-planta) measurements. At the conclusion of labeling experiments, the metabolism was quenched by a very brief rinse of the cotyledon surface with water prior to slicing off layers of the cotyledon to assess label uptake, metabolism, and heterogeneity (see Supplementary Data 1 for details). Slices were rapidly frozen in liquid nitrogen and stored at −80°C until extraction.
To investigate carbon turnover from lipids to carbohydrates, we substituted U-13C6 glucose with 13C3 glycerol (15 mM) as the sole source of carbon. 13C12 Sucrose (100 mM) was used as the carbon source for data presented in Figure 7. The salts and vitamins in all individual labelling experiments remained the same as above. The bottom slice of labeled seeds was used for metabolite extraction and measurements of isotopologue distribution (see Supplementary Data 1 for rationale).
RNA extraction and transcript analysis for verification of culturing system
Total RNA was extracted from soybean seed slices using the RNeasy Plant Mini Kit (Qiagen) according to the supplier’s instructions. cDNA was synthesized using the qScript cDNA SuperMix (Quantabio) from 1 μg total RNA previously treated with DNase I (Merck). For droplet generation, 20 μl of PCR reaction (cDNA, primers and the Bio-Rad ddPCR supermix) and 70 μl of droplet generation oil were transferred to the middle and to the bottom rows respectively of a DG8™ Cartridge before insertion into a QX200 Droplet Generator. The genes used and the primer sequences are included in Supplementary Table S4. Droplets were transferred to a 96-well plate for PCR amplification in a thermal cycler C1000 Touch™. The cycling protocol was 95 °C enzyme activation for 5 min followed by 40 cycles of a two-step cycling protocol of 95 C° for 30 seconds and Tm (Supplementary Table S4) for 1 min, then 4 °C for 5 min and 90 °C for 5min. Following PCR amplification, the plate containing the droplets was placed in a QX200 droplet reader. Droplet digital PCR (ddPCR) data was analysed with Bio-Rad QuantaSoft Analysis Pro Software. The Glycine max ATP synthase subunit 1 (Glyma12g02310) and SKIP16 (Glyma12g05510) were used as internal references (Hu et al., 2009).
Extraction of polar and non-polar metabolites
Metabolite extraction was carried out following the protocol described in Czajka et al. (2020)) and Kambhampati et al. (2019)) with a few modifications. Briefly, the stored samples were removed from −80°C and two metal beads were added to each tube along with 1 ml 7:3 methanol/chloroform (−20°C) and a PIPES (piperazine-N,N’-bis[2-ethanesulfonic acid]), norvaline, and ribitol mixed standard. Samples were kept on ice throughout extraction unless otherwise noted. Samples were pulverized using a ball mill at 28 Hz for 5 minutes or until fully ground. The mixtures were then incubated at −20°C for 3 hrs, with intermittent vortexing to ensure complete extraction. 500 μL of ddH2O (4°C) was added to each sample and vortexed vigorously before being centrifuged at 14,000 rpm at 4°C for 10 minutes during which the samples phase-separated. The upper aqueous phase containing water-soluble metabolites was transferred to a 1.5 mL eppindorf tube with a 0.45 μm centrifugal filter (Costar®, Corning Inc.) and spun at 14,000 rpm at 4°C for 2 mins. This solution was then transferred to glass vials (Agilent, Xpertek) for LC-MS/MS analysis to detect soluble sugars, free amino acids and sugar phosphates.
Quantification of proteins, lipids, and starch
3-5 mg of ground lyophilized tissue was subjected to liquid hydrolysis and protein was measured using amino acid compositional analysis as described in Kambhampati et al. (2019). In brief, 20 μL of 1 mM cell free 13C-labelled amino acid standard mix (Sigma) was added to the protein pellet and dried using a speed vacuum centrifuge. 50 μL of 4M methanesulphonic acid containing 0.2% tryptamine was added to this dried pellet and incubated at 110°C for 22 hours. Upon completion of hydrolysis, the samples were neutralized using 50 μL of 4M sodium hydroxide, briefly vortexed and dried. Upon drying, the samples were resuspended in 1 mL ultra pure water and vortexed to recover the hydrolyzed amino acids and then filtered using 0.45 μM centrifugal filters. Amino acids were detected using LC-MS/MS (described below) and quantified using isotopic dilution based on peak areas obtained from known concentrations of internal standards. The sum of the concentrations of all 20 amino acids, in milligrams, was used to establish the concentration of protein (Supplementary Table S5).
Analysis of lipid content was carried out according to an adapted version of the method described in Allen and Young (2013)) by converting total lipids into Fatty Acid Methyl Esters (FAMEs). In brief, freshly prepared 5% sulfuric acid:methanol (v/v) was added to ~20 mg of ground lyophilized tissue along with 25 ul 0.2% butylated hydroxytoluene (BHT) in methanol to prevent oxidation and two internal standards, triheptadecanoin and tripentadecanoin, before heating at 110°C for 3 hours, vortexing hourly. After cooling to room temperature, 0.9% NaCl (w/v) was added to each sample to quench the reaction. The FAMEs were then extracted using hexane and quantified by gas chromatography-flame ionization detection (GC-FID) using a DB23 column (30 m, 0.25-mm i.d., 0.25-μm film; J&W Scientific). The GC was operated in a split mode (30:1). The flame ionization detector was operated with a temperature of 250°C with an oven temperature ramp profile from 180°C to 260°C at a rate of 20°C min−1 followed by a hold time of 7 mins. Comparisons of peak areas to the two internal standards were used for quantification.
Starch measurements were performed on ~20 mg of ground lyophilized tissue, directly without prior extraction. Total starch content in cotyledons over reproductive development was determined, in triplicates, using the Megazyme starch assay kit (Megazyme International Ireland), using the AOAC Official Method 996.11 (Approved Methods of the AACC,(McCleary et al., 1997; McCleary et al., 2019)) modified to adjust the final assay volume for 96-well plate reader compatibility. Briefly, the ground lyophilized tissue was washed twice with 80% ethanol at 85°C prior to heating at 110°C for 10 min with DMSO. The samples were then treated with α-amylase at 110°C for 12 minutes (vortexing every 4 min) followed by amyloglucosidase at 50°C for 1 hour. The samples were then centrifuged, supernatant collected, and incubated with the GOPOD reagent at 50°C for 20 min. The absorbace at 510 nm was measured using a spectrophotometer, and the starch content was determined by comparison with a standard curve generated using a serial dilution of starch standards treated the same way as biological samples. Quantities of all biomass component measurements presented in Figure 2 are provided as Supplementary Table S6.
Quantification of soluble sugars, amino acids, and sugar phosphates using HPLC-MS/MS
Sugars and sugar phosphates were analyzed from the water-soluble fraction using a Shimadzu (UFLCXR) HPLC system connected to an AB Sciex triple quadrupole MS equipped with Turbo V™ electrospray ionization (ESI) source using the method described in Czajka et al. (2020)). Negative ion mode was used to monitor sugar and sugar phosphate fragments. A 5 μl sample was injected on the Infinity Lab poroshell 120 Z-HILIC column (2.7 μm, 100 x 2.1 mm; Agilent technologies, Santa Clara, CA, USA) and the metabolites were eluted with an increasing gradient of acetonitrile: 10 mM ammonium acetate (90:10 v/v) and 5 μm medronic acid, pH 9.0 (A) and 10 mM ammonium acetate in water, pH 9.0 (B). The flow rate was 0.25 mL/min. Sugars and sugar phosphates were separated using a binary gradient of 95-70% B over 8 minutes then to 50% B over the next 4 min followed by a hold at 25% B for 1.5 min. The gradient was then decreased to 30% B over 0.5 min followed by a hold for 1 min before returning to 95% B to re-equilibrating the column for 6 min. The HPLC eluent was introduced into an electrospray ionization source with the following conditions: ion spray voltage, 4.5 kV (ESI-); ion source temperature, 550°C; source gas 1, 45 psi; source gas 2, 40 psi; curtain gas, 35 psi and entrance potential, 10. Ions were detected and monitored using a targeted MRM approach with the parameters optimized by direct infusions, provided in supplementary Table S7, for accurate quantification. The value for entrance potential was default (−10) for all analytes. For absolute quantification, data were analyzed using the quantitation wizard available in Analyst (v. 1.6.2) software (AB SCIEX, Concord, Canada). Metabolite concentrations were calculated based on a calibration curves. Recoveries were assessed using ribitol and PIPES as internal standards for sugars and sugar phosphates, respectively.
Amino acids were measured using the same instrumentation and column as described above, with the following changes in mobile phases, gradient and ionization conditions; Mobile phase A consisted of 20mM ammonium formate in water, pH 3.0, and B was composed of 90% acetonitrile and 10% water with a final concentration of 20 mM ammonium formate, pH 3. 3 μL from each sample were injected and a flow rate of 0.25 μL was used for separation of amino acids on the HPLC column. A binary gradient composed of 100-90% B over 2 minutes, 90-50% B over the next 6 minutes followed by returning to 100% B over 30 seconds and re-equilibration of 5.5 minutes was used to separate the analytes. The HPLC eluent was introduced into an electrospray ionization source with the following conditions: ion spray voltage, 4.5 kV (ESI+ and ESI-); ion source temperature, 400°C; source gas 1, 45; source gas 2, 40; curtain gas, 35 and entrance potential, 10. Ions were detected and monitored using a targeted MRM approach using parameters included within Supplementary Table S7. Data were analyzed similar to sugars and sugar phosphates described above, except novaline was used as an internal standard to assess recoveries. All statistical analysis and data visualization were performed using Microsoft Excel (2013) or R programming language (R CoreTeam, 2013) using base functions and the package ggplot2 (Wickham, 2016).
Quantification of Isotopologue abundance and average labeling
The LC-MS/MS conditions used for label detection are identical to the ones described above with the exception of MRM transitions used (Supplementary Table S8) that were selected based on Kappelmann et al. (2017). Peaks were manually integrated and the natural abundance was corrected using the R package IsoCorrectoR (Heinrich et al., 2018). Fractional enrichment of the corrected isotopologues (M0-Mn) obtained from IsoCorrectoR was used to calculate average labeling. Average labeling was calculated as described in Buescher et al. (2015) using the following equation; where i denotes the mass isotopologue, n is the number of possible 13C carbons, and S is the fraction of the labeled isotopologue. For detailed calculations see Supplementary Table S9.
PEPCK enzyme activity assay
PEPCK enzyme activity was measured using the method of Walker et al. (1999) detailed on protocols.org (Osorio et al., 2014). Briefly, crude protein was extracted from 100-250mg FW of seed tissue in a buffer containing 0.5M bicine-KOH (pH 9.0), 0.2M KCl, 3mM EDTA, 5% (w/v) PEG-4000, 25mM DTT, and 0.4% bovine serum albumin. The extract was centrifuged for 20 minutes at 13,000 x g at 4°C, and the supernatant was added to a buffer containing 0.5M bicine-KOH (pH 9.0), 3mM EDTA, 55% (w/v) PEG-4000, and 25mM DTT, then incubated for 10 minutes on ice and centrifuged at 13,000 x g at 4°C for 20 minutes. The supernatant was discarded, and the pellet was resuspended in 10mM bicine-KOH (pH 9.0) containing 25mM DTT. PEPCK activity was measured in the direction of the carboxylation reaction by coupling the reaction with malate dehydrogenase (EC 1.1.1.37; Sigma Aldrich 10127914001) and following the oxidation of NADH at 340nm at room temperature using a spectrophotometer (SpectraMax M2e, Molecular Devices). Total protein was measured using the protein extract for PEPCK activity using Bradford reagent (Millipore Sigma; Cat: B6916) and a standard curve using commercial bovine serum albumin standards (Thermo Fisher, cat: 23208).
Supplementary material
Supplementary data 1: Establishing and assessing short time pulse labeling conditions for non-perturbed in planta temporal assessment of seed development
Supplementary Figure S1: 13C6 glucose labeling using cotyledons of R7 seeds for establishing the culturing system.
Supplementary Table S1: Composition of seed coat exudate, quantities represented in nmol per seed
Supplementary Table S2: Total pool size (accurate quantities) of the metabolites detected via LC-MS/MS
Supplementary Table S3: Temporal changes in carbon conversion efficiency and loss of CO2 as PEPCK
Supplementary Table S4: Primers used for droplet digital PCR and their annealing temperatures
Supplementary Table S5: Amino acid concentration (mg seed−1) from hydrolyzed protein at different developmental stages
Supplementary Table S6: Biomass component measurements for soybean seed developmental stages
Supplementary Table S7: MRM parameters used for absolute quantification
Supplementary Table S8: MRM parameters used for isotope labeling experiments
Supplementary Table S9: Calculations for isotopologue distribution and average labeling
Acknowledgements
The authors would like to thank the director of proteomics and mass spectrometry (PMSF) facility at the Donald Danforth Plant Science Center (DDPSC), Dr. Bradley S. Evans for his support and advice throughout this work. In addition, the authors acknowledge the Plant Growth Facility (PGF) at DDPSC for their help with plant growth and maintenance. The authors declare no conflicts of interest.
Footnotes
Funding information This work was supported by the United States Department of Agriculture, Agricultural Research Service, the United States Department of Agriculture, National Institute of Food and Agriculture through award number USDA-NIFA #2017-67013-26156, #2016-67013-24585, and the Multistate Research Project NC1203, the United Soybean Board (#1820-152-0134), National Institutes of Health (U01 CA235508), and the National Science Foundation (NSF-1812235). Support for the acquisition of the 6500 QTRAP LC–MS/MS was also provided by the National Science Foundation (NSF-DBI #1427621).