Seed growth and accumulation of storage products relies on the delivery of sucrose from the maternal to the filial tissues. The transport route is hidden inside the seed and has never been visualized in vivo. Our approach, based on high-field nuclear magnetic resonance and a custom made 13C/1H double resonant coil, allows the non-invasive imaging and monitoring of sucrose allocation within the seed. The new technique visualizes the main stream of sucrose and determines its velocity during the grain filling in barley (Hordeum vulgare L.). Quantifiable dynamic images are provided, which allow observing movement of 13C-sucrose at a sub-millimetre level of resolution. The analysis of genetically modified barley grains (Jekyll transgenic lines, seg8 and Risø13 mutants) demonstrated that sucrose release via the nucellar projection towards the endosperm provides an essential mean for the control of seed growth by maternal organism. The sucrose allocation was further determined by structural and metabolic features of endosperm. Sucrose monitoring was integrated with an in silico flux balance analysis, representing a powerful platform for non-invasive study of seed filling in crops.
Seeds form the basis of much of the human and domesticated animal diet, as well as representing an important feedstock for industry. The accumulation of starch, proteins and lipids in the endosperm and embryo requires a continuous supply of nutrients from the maternal plant, much of which is in the form of sucrose. Seed filling parameters are associated with agronomic traits. Thus, selection of genotypes with high seed filling rates has been proposed as a promising strategy for raising crop productivity. The mechanisms underlying sucrose delivery to the developing seed are a prominent topic of crop improvement research (Bewley et al., 2006). A number of the genes involved in sucrose transport have been identified (Lalonde et al., 2004; Braun and Slewinski, 2009; Kühn and Grof, 2010), and their contribution to seed filling has been evaluated (Weber et al., 1997; Matsukura et al., 2000; Baud et al., 2005). However, control of sucrose allocation, i.e. diversion of sucrose into the various seed tissues, appears very complex. Sucrose transporters are tightly regulated at different levels, and transporters functioning to efflux sucrose remain to be identified (Kühn and Grof, 2010). Cellular specialization, particularly in the context of the so-called transfer cells (McCurdy et al., 2008; Gomez et al., 2009), was recognised as an important component of the seed filling process (Thompson et al., 2001; Royo et al., 2007). Several models for post-phloem translocation have been developed (Patrick and Offler, 2001; Zhou et al., 2009), but an understanding of how seed filling functions in living seed calls for appropriate in vivo studies. The structure of the seed impedes direct access to the site of the interaction between the maternal and the filial tissues. Any invasive method is linked with risks of artificial changes in both the in vivo metabolite distribution and enzymatic activities.
Currently, the method that has been most fruitful in non-invasive studies of solute transport in plants is nuclear magnetic resonance (NMR) imaging. Since the introduction of non-invasive flow measurements, great progress has been made towards the understanding of water (Van As and Schaafsma, 1984; Van As, 2007) and sucrose dynamic in vascular tissue (Verscht et al., 1998; Szimtenings et al., 2003). Initial imaging of water movement in seeds was achieved about 20 years ago (Jenner et al., 1988), but only few non-invasive studies on sucrose are available which in fact do not access its movement (Ischida et al., 2000). The slow flow velocity and extremely low flowing volume per unit of cross-section has made monitoring of sucrose on sub-millimetre scale a technically challenging enterprise. The elaboration of non-invasive means to study the post-phloem allocation of sucrose within seed tissues is thus an important research priority.
Radioactive 14C labelling has long been used as a means of tracking the uptake and fate of sucrose (Fisher and Gifford, 1986). A disadvantage of this technology, however, is that the signal is not strong enough to be detected in intact tissue, which means that these assays have to be destructive. This problem does not arise when the 11C isotope is used, as this nucleus emits a positron which annihilates with an electron to high-energy γ-radiation (Minchin and Thorpe, 2003). However, the 11C half life is only 20 min, thereby limiting its utility in biological experiments (Jahnke et al., 2009). NMR provides an alternative in vivo detection platform (Ratcliffe et al., 2001; Koeckenberger et al., 2004). Using 1H NMR, which exploits the signal emitted by protons associated with carbon nuclei, sucrose can be detected, imaged and quantified (Tse et al., 1996; Melkus et al., 2009). NMR can also directly measure 13C sucrose (Kalusche et al., 1999). When 13C labelled molecules are fed to a plant (Ekman et al., 2008), not only the sensitivity of the 13C NMR imaging is increased, but also some information regarding metabolite uptake and distribution can be recovered (Heidenreich et al., 1998; Kalusche et al., 1999; Roberts, 2000). This study aims to elucidate the sugar allocation pathway and to characterize the sucrose movement in the intact living seed.
Barley is recognized as an appropriate model for the improvement of seed yield in cereals for various reasons (Sreenivasulu et al., 2010a), not the least of which is that it is an important crop in its own right. Here, time-resolved maps of sugar distribution were generated using direct and inverse 13C/1H NMR imaging methods, allowing the monitoring of sugar movement inside the developing seed.
Experimental design for the in vivo sucrose monitoring using direct and inverse 13C/1H NMR imaging of the individual caryopsis
As a first step, 13C sucrose uptake and distribution within the spike was traced using conventional (invasive) procedures. The uptake into individual grains was measured after 16 h of feeding with 13C sucrose via the detached stem. The absolute accumulation of 13C in the individual caryopses was most regular within the middle of the spike (Figure 1a). The relative abundance of 13C in individual caryopses was at a level of 4.02 ± 0.36% of total carbon, and there was no evidence of any gradient along the spike. When dissecting individual caryopses, about 86% of the 13C label was retained in the starchy endosperm, while 14% was detected in the pericarp and a trace in the embryo.
To acquire non-destructive structural (1H NMR) and functional (13C NMR) images during a single experiment, a double-resonant 13C/1H NMR coil was constructed that matched with the size of the caryopsis. The coil was fixed to the central portion of an intact spike, enclosing one caryopsis (Figure 1a). The NMR probe head including the coil and the spike was positioned in a 17.6 T vertical magnet equipped with actively shielded imaging gradients (see Methods for details). This setup allowed the acquisition of anatomical 1H and dynamic 13C images without moving the object. Seed anatomy was captured by 1H NMR imaging using a multi-slice spin echo sequence, while both direct and inverse 13C acquisition were applied for the imaging of 13C (Figures 1b, c). The direct imaging of 13C involved excitation with a pulse at the 13C Larmor-frequency, with the signal being acquired on the same channel (Figure 1b). The signal-to-noise ratio (SNR) of this sequence was improved by 500 ms proton saturation prior to 13C excitation, leading to signal amplification through the Nuclear Overhauser Effect. Further, SNR improvement was obtained by 1H decoupling, a procedure that suppresses the 1J coupling between 1H and 13C nuclei, thereby merging the 1J-splitting in the 13C spectrum. Inverse detection relies on the excitation of protons, followed by the use of specific frequency pulses and magnetic field gradients to select and acquire the protons associated with the 13C nuclei. They are generally more sensitive than direct acquisition methods, because the gyromagnetic ratio γ of the 1H nucleus is four-times larger than that of the 13C nucleus and NMR sensitivity is approximately proportional to γ2 (Gruetter et al., 2003). We applied the gradient-enhanced heteronuclear multiple quantum coherence (geHMQC) method for the inverse detection of 13C (Hurd and John, 1991), combined with additional water suppression, phase encoding for localization and 13C decoupling at acquisition (Figure 1c) to improve the SNR of the spectra. Typical spectra acquired by the two 13C detection schemes are illustrated in Figures 1d–f. Although both in vivo spectra showed line broadening owing to magnetic susceptibility effects (Figure 1f, g), the sucrose signal clearly remained resolvable. We also checked the 13C-sucrose uptake of caryopses by mass spectrometry-based analysis. During the time course of feeding, there was an increase in the steady state level of 13C labelled sucrose as well as the 13C label in starch (data not shown).
Dynamic imaging of 13C sucrose stream within the intact seed
The kinetics of 13C uptake was monitored by 13C/1H NMR imaging in individual grains attached to the spike at the onset of their storage activity. The distribution of label was visualized in a cross-section of the spike (Figure 2) fed continuously with 13C sucrose. The colour-coded level of 13C sucrose was imaged as a 2D plot every 30 min over a 12-h period, and then animated to give an time-compressed visualization of 13C sucrose allocation within the caryopsis (Movie S1). The NMR signal became apparent when the 13C sucrose concentration exceeded 4 μmol per g dry weight of caryopsis (mass spectrometry analysis). The label appeared first in the region of the vascular bundle, clearly identifying the crease vein as the main sucrose entry point. The sucrose then moved via the nucellar projection towards the endosperm cavity, where it was enriched (colour shift from blue to red). Thereafter, the signal increased in the endosperm, appearing first in the transfer cells and the attached starchy parenchyma, and decreasing gradually towards the periphery. There was no evidence of any inflow of sucrose into the endosperm from either the dorsal pericarp or the circumferential nucellar epidermis. These data show that firstly there is a preferential allocation of sucrose towards the endosperm cavity, and secondly that there is a localized accumulation of sucrose in the apoplastic space. The enrichment of sucrose and its further pattern of expansion were also demonstrated by the spatial progression of isolines over time (Figure 3a). These lines of constant signal were extracted from the NMR data using a threshold well above the noise level. The velocity of the labelling front in dorso-ventral direction reached 145 μm/h, but only 72 μm/h within the lateral regions of pericarp. The sucrose allocation route remained unchanged until the late developmental stage (Figure S1). The 13C sucrose stream within the caryopsis was interrupted, and its gradients disappeared when the caryopsis was detached from the spike (data not shown).
Structural determinants for sucrose allocation along the maternal-filial pathway
Cell morphology, thickness of the cuticle and other physical barriers may affect the distribution of sucrose. On the ventral side, the crease vein is surrounded by an arc of small parenchymatous cells which appear to be cuticularized (Figure 3c). These structures could hamper the lateral and ventral movement of sucrose. The elongated cells of the nucellar projection are directed towards the endosperm, aiming the sucrose flow to the filial tissue (Figure 3d). Multiple symplastic junctions, large intercellular spaces and cell wall ingrowths are found along the transport route (Figure 3e). As a result, sucrose is more likely to move in a dorsal than in a lateral or ventral direction (see Movie S1). The high vacuolization of cells within the nucellar projection allows for the transient accumulation of sucrose. Autolyzing efflux cells release their complete cell content towards the liquid apoplast called endosperm cavity (Figure 3f, g). The highest level of 13C label within the caryopsis was found in the endosperm cavity and proximal regions. Sucrose entered the filial side via the endosperm transfer cells (ETCs, Figure 3h). These specialized cells permit the passage of metabolites (Offler et al., 2002) and do not possess features of a storage tissue. This is in contrast to the endosperm parenchyma cells that are involved in the active accumulation of starch (Figure 3i). A thick cuticular layer (Figure 3j) borders the pericarp and encloses the whole endosperm except for the crease region. We conclude that channelling of the sucrose stream towards the endosperm rely on the cellular architecture of the crease gate.
Differential gene expression reflects the functional specialization of the crease region
A comparison of gene expression between the crease and dorsal regions of the pericarp (Figure S2a) was performed using a seed-specific 12 K macro array (Sreenivasulu et al., 2006). Applying a minimum threshold of a threefold difference in expression level, 614 genes were identified as being differentially expressed. The largest classes of genes up-regulated in the crease region were those associated with programmed cell death (PCD), transport and metabolism (Table S1). The most prominently up-regulated was Jekyll gene (Table 1), which plays a pivotal role in the differentiation and fate of the nucellar projection and reach the maximum expression levels in strongly vacuolated and deteriorating efflux cells (Radchuk et al., 2006). Other genes potentially involved in the PCD acquisition (Hatsugai et al., 2006) such as vacuolar processing enzymes VPE2a and VPE2b (Radchuk et al., 2010), metacaspase 9 and a nucellin gene encoding an aspartic protease (Chen and Foolad, 1997) were up-regulated in crease region. The genes encoding enzymes of cell wall degradation such as endo-1,4-β-glucanase (CEL1, CEL3, CEL9) and β-D-xylosidase (BXL1, BXL2) were activated. As indication of the strong sucrose transport competence of crease, the transcript levels of sucrose transporter 1 (SUT1) and plasma membrane H+-ATPase genes were elevated. In situ hybridization localized the SUT1 expression to the nucellar projection and the region surrounding the vein (Figure S2). Genes involved in sucrose inter-conversion, as for example vacuolar invertase (VIN1), glycolytic pathway and genes related to amino acid metabolism were up-regulated in crease (Table 1). The transcriptional activation of aquaporins localized on the plasma membrane (PIP1;3, NIPs) and the tonoplast (γTIP and δTIP) in crease region indicates higher rates of nutrient transfer and/or unloading through the formation of osmotic gradients. Among transcription factors, several gene family members of MADS-box TFs were found to be preferentially expressed in the crease region (Table S1). In rice, OsMADS6 suppression leads to dramatic alteration in the nutrient content of grains, suggesting that OsMADS6 might have a critical role in nutrition (Zhang et al., 2010). In essence, differential gene expression reflects the functional specialization of the crease region for delivery of assimilates and necessarily involves cell death.
Table 1. Partial list of the genes that are up-regulated at least threefold in the crease region
Gene Identification [Species]
Fold change in expression
Full list of genes up- and down-regulated in the crease region in comparison with the dorsal region of a barley grain is shown in Table S1.
Metabolite and enzyme profiles evidences differential sucrose metabolism in the maternal and filial parts of caryopsis
To compare the sucrose metabolism at the maternal and the filial site of caryopsis, we measured metabolites and enzymes in dissected pericarp in comparison with endosperm. The inter-tissue differences are summarized in Figure 4. In pericarp, cell wall-bound invertase (INV) was the dominant sucrose cleaving enzyme, but sucrose synthase (SuSy) in the endosperm. The predominant hexose in the pericarp was fructose (produced by invertase), while in the endosperm it was UDP glucose (produced by SuSy). Little starch was present in the pericarp, especially compared to the levels present in the endosperm. Congruently, the endosperm contained appreciable levels of the starch precursor ADP glucose. The level of several glycolytic intermediates was higher in the endosperm than in the pericarp, which together with the more elevated levels of alanine and succinate are thought to indicate an accelerated rate of glycolysis resulting from the hypoxic status of endosperm cells (Rolletschek et al., 2004). The hexose-to-sucrose ratio in the endosperm cavity (apoplastic liquid between maternal and filial part) was 0.08 ± 0.03, indicating that sucrose represents the bulk of the carbohydrates transferred from the maternal tissue via the nucellar projection into the filial tissue.
Defective nucellar projection compromises sucrose flow to the endosperm
The role of nucellar projection in sucrose transport was further investigated by the analysis of transgenic plants with defective nucellar projection owing to down-regulation of the Jekyll gene (Radchuk et al., 2006). We applied 13C/1H NMR to measure the effect of Jekyll repression on the allocation of sucrose in transgenic plants. The velocity of sucrose movement along the crease gate was 83 μm/h (or 60% of the WT level) in transgenic line 61, 56 μm/h (40%) in line 18 and nearly broken in line 91. This coincides with the degree of Jekyll down-regulation (60%, 70% and 80%, correspondingly; Radchuk et al., 2006). The spatial distribution of 13C within the caryopsis was not altered in lines 61 and 18 (data not shown), but drastically changed in line 91 (Figure 5f), in which the nucellar projection is represented by a disorganized tissue (Figure 5a, c). Sucrose is barely released from the nucellar projection and is accumulated in its expanded cells, whereas the endosperm is underdeveloped. Thus, the suppression of Jekyll expression clearly decelerate the sucrose release into the endosperm cavity, with the result that sucrose availability to the endosperm is reduced and the accumulation of starch is decreased (Figure 5b, d).
We compared gene expression patterns of WT crease region and endosperm fraction with those of the transgenic Jekyll down-regulated line 91. The severely affected specification of cells in crease region and disturbed storage process in endosperm of transgenic caryopses were reflected in the differential expression profile of regulatory and PCD-related genes, as well as genes involved in sucrose transport and metabolism (Figure 5g,Table S3). Genes encoding fructan-6-fructosyltransferases were activated in the transgenic crease region, indicating a switch from high sucrose to high hexose. In a concerted action, enolases from the glycolytic pathway were substantially elevated indicating a more active metabolism within the massively proliferated nucellar projection (Figure 5c, d). In the transgenic endosperm, the SUT1 expression was down-regulated. Down-regulated expression of susy2 was obvious in both transgenic endosperm and crease region. Because SuSy is the key sink building enzyme in endosperm, a decrease in the starch content might be expected. Thus, key factors affecting endosperm storage events in the transgenic line were linked to sucrose transport and metabolism. This explains the reduced size of the mature caryopses and their decreased starch content as a characteristic phenotype of these transgenic plants (Radchuk et al., 2006).
Structural and metabolic alterations in endosperm are reflected on allocation pattern and velocity of sucrose stream
We have analysed the barley shrunken endosperm mutant seg8 (Ramage and Crandall, 1981; Sreenivasulu et al., 2010b). The size of the starchy endosperm adjacent to the nucellar projection was heavily reduced (Figure 6a, b). There were noticeable delay and irregularity in the cell wall formation in this region during early development, leading to the formation of abnormal ETCs (Figure S3). Most probably, these aberrant cells themselves metabolize sucrose to starch, rather than forward it to the adjacent endosperm cells as do WT ETCs. The starchy endosperm within the crease region was consequently reduced to just a few cell layers (Figure 6g). The lateral wings of the seg8 endosperm showed properly developed ETCs and maintained their regular form and function.
The in vivo NMR experiments showed that the 13C sucrose concentration gradient along the primary transport route was sharpened in the seg8 mutant (Figure 6e, f). The sucrose spreads rather in lateral direction, pointing to higher flow resistance in central endosperm where ETCs/endosperm is aberrant. The total level of 13C sucrose in the seg8 endosperm was decreased as compared to the WT (Figure 6h, i). The transcript analysis of the seg8 endosperm has revealed down-regulation of genes involved in starch synthesis including sucrose synthase, ADP glucose pyrophosphorylase, starch synthase, starch branching and debranching enzymes (Table S4). In contrast, the higher concentrations of both sucrose and ADP glucose were accumulated in seg8 pericarp, as compared to the WT (Figure 6j, k). Congruently, higher starch accumulation was visible in ventral pericarp of the mutant already during early development (Figure 6l, m). The relocation and reduction in sucrose transfer between the maternal and filial tissues is readily understandable owing to the abnormal structure of the filial side of the transport route.
The Risø13 mutant also accumulates little starch in its endosperm, but in contrast to seg8, appears to be morphologically normal. The mutation is known to cause a lesion in the delivery of ADP glucose from the cytosol to the site of starch synthesis in the plastid (Patron et al., 2004). The 13C signal intensity and its distribution pattern in Risø13 were identical to those of the WT (data not shown), but the velocity of sucrose relocation from maternal to the filial part was reduced to 68% of WT level. This reduction is likely connected to the lowered metabolic demand of the Risø13 mutant endosperm.
Simulation of the endosperm metabolism in Jekyll down-regulated plants and Risø13 mutant plants based on flux balance analysis
13C/1H NMR monitoring visualized maternal and filial effects on sugar allocation within the living seed, and here we aim to elucidate the biochemistry behind the images. The flux balance analysis (FBA) in the WT, the transgenic Jekyll down-regulated plants and the Risø13 mutant was performed using the model (including biomass composition) established earlier (Grafahrend-Belau et al., 2009a), and by constraining the sucrose uptake rates at representative values derived from in vivo13C NMR measurements. A detailed description of all modelling constraints is given in the Methods section, the resulting flux values are given in Table S5. In the following, the modelling results are described and the changes in the metabolic behaviour in comparison with the WT are specified.
In the Risø13 mutant flux through the cytosolic isoform of AGPase was greatly decreased (Figure 7). The plastidial isoform became the major source of ADP glucose synthesis, depending on the import of hexose phosphates into the plastid. In contrast to WT, glycolytic flux was restricted to the ATP-dependent reactions of phosphofructokinase (PFK) and pyruvate kinase (PK). At the same time, the phosphofructophosphatase (PFP) and pyruvate-phosphate dikinase (PPDK) bypass reactions acted in the direction of PPi regeneration to compensate for the reduced cytosolic PPi synthesis resulting from a low flux through cytosolic AGPase. Because of the reduction in storage product synthesis, surplus carbon needed to be removed from the system, and this was achieved by the accumulation of UDP glucose and ADP glucose (applied as modelling constraint, see Methods for more detail), as well as by the excretion of CO2 (mostly respiration). In the mutant, CO2 emission was raised by 40% over the WT level. Taken together, the severe reduction in the availability of the plastidial ADP glucose transporter resulted in a strong decrease in starch synthesis and biomass production, as well as inducing a number of metabolic changes in the starch synthesis pathway.
In the Jekyll down-regulated plant, the metabolic flux patterns differed greatly with respect to their flux values, but not with respect to their pathway utilization in comparison with WT (Figure S4). The reduction in sucrose supply to the endosperm resulted in a large decrease in the deduced rate of storage products synthesis, leading to a > 50% decline in the endosperm growth rate. Sucrose degradation was largely restricted to the SuSy pathway. Synthesis of ADP glucose was predominantly catalysed by the cytosolic isoform of AGPase. Glycolytic flux was restricted to the PFP and PPDK PPi-utilizing bypass, while flux through the TCA cycle was high. Induced flux through the anaplerotic reactions of cytosolic Ala-aminotransferase, mitochondrial Asp-aminotransferase, cytosolic and mitochondrial malate dehydrogenase resulted in the replenishment of the TCA cycle intermediates, thereby allowing some compensation for the loss of amino acid precursors such as malate, oxaloacetate and 2-oxoglutarate, which are normally required for storage product synthesis. Seed storage metabolism was fuelled by carbohydrates allocated by the action of phosphoenolpyruvate carboxylase.
Advantages of the 13C/1H NMR-based imaging method for characterizing seed filling
To track assimilate allocation in individual seeds, the non-invasive tool should not only detect the metabolite but also allow adequate spatial and temporal resolution during long-term monitoring. Structural parameters of seeds can be extensively analysed using magnetic resonance imaging (MRI) and X-ray tomography (Koeckenberger et al., 2004; Friis et al., 2007). The visualization of metabolite distribution and dynamic changes, so-called functional imaging (Van As, 2007; Frommer et al., 2009), has remained technically challenging and is hardly available for seed biology. The speed with which solutes move within the plant vasculature is low (Mullendore et al., 2010), and the speed of their dispersion outside the vasculature is expected to be much slower (Jenner et al., 1988; Patrick and Offler, 2001). Flow encoded NMR measurements (Van As, 2007) are difficult to perform when the velocities are in the μm/h range, because of the need for long and strong encoding gradients and a long gradient separation time which leads to immense decrease in SNR. Such sequences are effective, where velocities are on the scale of thousands of μm/h (Szimtenings et al., 2003). The fluorescence resonance energy transfer technology (FRET) was successfully used for sub-cellular real-time carbohydrate monitoring (Niittylae et al., 2009), but its application to seed is rather restricted. Alternatively, NMR and positron emission tomography (PET) allow the development of a ‘molecular imaging’ technique to look through opaque tissue non-invasively, even if the diameter of the sample is large (Koeckenberger et al., 2004; Phelps, 2004). PET has shown great potential for in planta analysis because of to its high sensitivity (Babst et al., 2005; Jahnke et al., 2009), but when 11C is chosen as the monitored isotope, the spatial resolution of the assay is limited to approximately 1.4 mm (Phelps et al., 1975). Thus, PET is more convenient to study long-distance translocations in plants (Schwachtje et al., 2006; Jahnke et al., 2009). NMR - and especially 13C NMR - is not as sensitive as PET (Metzler et al., 1995; Melkus et al., 2009). Nevertheless, in our hands, 13C NMR delivered a fivefold higher in-plane resolution than PET and allowed dynamic observations. To achieve this quality, 13C detection methods such as the geHMQC sequence (Hurd and John, 1991) were combined with an optimally adapted double resonant detector coil and a high magnetic field strength. The metabolic images were captured on a time scale of one hour for the direct 13C measurements, and 30 min for the inverse detection scheme. Particularly for the non-invasive investigation of seed, for which the spatial scale is measured in millimetres, the 13C NMR method was thus able to achieve a sufficient level of resolution within a reasonable measurement period. Using PET, it is not possible to determine with which specific molecule the decaying 11C nucleus is associated. The present method, by contrast, not only detects 13C labelled sucrose, but also has the potential to identify other labelled metabolites (own unpublished data).
Finally, the 13C/1H NMR method allows for the straightforward co-registration of the structural and the metabolite images, which enables the exact localization of metabolites within a tissue. Based on these features, it has therefore become, for the first time, possible to visualize sugar movement within the seed.
Implications of non-invasive sucrose imaging on the modelling of grain filling
Different dyes, fluorescence or isotope-labelled substances were applied to trace transport sucrose routes (Wang et al., 1994;Fisher and Cash-Clark, 2000; Stadler et al., 2005). The extent to which models mirror the in vivo sucrose exchange between the maternal and filial organism can always be questioned. The NMR approach directly visualized sugar movement and highlighted its transport route.
The main gateway for sucrose inflow during the grain filling is represented by the crease vein, placento-chalasal tissues, nucellar projection, the endosperm cavity and the ETC. These tissues work in concert for channelling of sucrose from the site of sucrose delivery to its utilization. Multiple symplastical connections may help to increase the conductivity along the transport route (Fisher and Cash-Clark, 2000) over that of other pericarp tissues. A characteristic cell differentiation gradient was detected within the nucellar projection, a tissue in which the cells were gradually enlarged, stretched in the direction of solute flow and collapsed in the region abutting the endosperm cavity. This process involves concomitant expression of the sucrose transporter SUT1 (Figure S2) and SUT2 (Weschke et al., 2000; Endler et al., 2006), which prompts accumulation of sucrose in cells (Sivitz et al., 2005). Nucellar projection does not store starch, and thus the accumulated sucrose is released into the endosperm cavity during the autolysis of “efflux cells”. Within the genes which arrange functional divergence of the crease gate, the group of PCD-related genes was pronounced (Table 1). The most up-regulated gene Jekyll is closely related to cell differentiation and cell death (Radchuk et al., 2006). When the fate/autolysis of efflux cells is disturbed owing to Jekyll repression, sucrose release towards the endosperm is decreased. Results of dynamic sucrose imaging emphasize the function of efflux cells at the crease gate as essential mechanism controlling sucrose delivery. Rather similar patterns of sucrose release were suggested for wheat (Wang et al., 1994; Domínguez et al., 2001), rice (Matsuo and Hoshikawa, 1993) and maize (Kladnik et al., 2004). Despite the remarkable differences in the seed anatomy across a range of species (reviewed by Patrick and Offler, 2001), seed plants seem to use a well-conserved cellular adaptation to fulfil the efflux function, namely controlled cell death.
We further demonstrate that the entering of sucrose into the endosperm occurs predominantly in a radial outward direction from the endosperm cavity and continues with the same velocity centrifugally throughout the endosperm (Figure 2, Movie S1). Sucrose forms a steep and continuous gradient with its peak at the ETCs, and its trough at the periphery of the endosperm. When ETC identity is changed as in seg8 mutant, the sucrose stream is reduced and redirected. In wheat, ETCs and sub-aleurone cells also express SUT1 and might contribute to sucrose uptake (Patrick and Offler, 2001). In rice, ETC morphology is not pronounced (Furbank et al., 2001) and the expression of SUT1, which is responsible for sucrose uptake into the grain (Scofield et al., 2002), is localized to the aleurone (Ishimaru et al., 2001). Sucrose is thought to move centripetally once it has entered the endosperm (Lim et al., 2006). The ETC of maize endosperm are sited opposite the pedicel (Kladnik et al., 2004), and sucrose might move from there straight forward into the endosperm. We believe that application of 13C/1H imaging to other crops will provide new insights into general mechanisms governing the assimilate stream in living seeds, and thereby facilitate crop improvement.
Non-invasive sucrose imaging uncovers maternal/filial interactions during seed filling
The quantitative nature of NMR imaging allows connecting in vivo analytics and in silico simulation, which has been shown to be a powerful approach (Di Ventura et al., 2006). To assess the complex biochemical machinery behind the dynamic images, we applied the FBA operating with 257 biochemical and transport reactions across the four compartments cytosol, mitochondrium, plastid and extracellular space in the barley caryopsis (Grafahrend-Belau et al., 2009a).
The results of the FBA simulation suggest that the down-regulation of Jekyll expression causes a marked decrease in biomass (starch) accumulation and changes in absolute fluxes, respectively. However, the ratio of internal fluxes remains almost constant, indicating that there is no major rearrangements of the metabolic network in endosperm. Similar proportional changes were found in Arabidopsis (Williams et al., 2008) and other species, and point to the network stability as a general feature of carbon metabolism in plants. The reduction in starch synthesis further implicates that caryopses with a defective nucellar projection are likely to develop lower sink strength. Accordingly, the expression of SUT1 and susy2 genes is reduced in Jekyll down-regulated endosperm, which is in accordance with decreased starch content and smaller size of mature caryopses in transgenic plants (Radchuk et al., 2006). We propose that sucrose release via the nucellar projection towards the endosperm provides an essential mean for the maternal tissue to impose control on the growth of the seed and the accumulation within it of storage products. In the seg8 mutant, the failure of ETC differentiation might result in higher resistance to sucrose flow. The 13C/1H NMR imaging visualized aberrant sucrose allocation and decreased uptake into the caryopsis. Accordingly, the transcript level of genes involved in starch biosynthesis was reduced. These data emphasize the importance of the transport pathway for the control over sucrose flow and regulation of sink strength (Minchin and Thorpe, 1996).
The sucrose allocation in the Risø13 mutant did not reveal any spatial alterations, but decreased velocity of sucrose stream through the crease gate, as compared to the WT. The FBA applied to the Risø13 mutant demonstrates distinct shifts in primary metabolism (e.g. PPi cycling), thereby generating novel targets for more detailed metabolic studies. The simulated flux to starch was substantially reduced, corresponding to the observed failure in starch accumulation in this mutant (Patron et al., 2004). The consequent fall in carbon demand is expected to reduce sink activity (Marcelis, 1996). The kinetics of sucrose allocation from the pericarp to the endosperm is correspondingly altered in the Risø13 mutant. Overall, the modelling data confirmed that the cooperative action of enzymes controlling the partitioning of sucrose into starch regulates sink strength, and that this finally determines seed yield (Koch, 2004; Frommer and Sonnewald, 2010).
We conclude that non-invasive 13C/1H NMR imaging is an appropriate platform for the analysis of structural, metabolic or genetic effects on seed filling. The post-phloem allocation of sucrose can now be investigated at the sub-millimetre scale, filling the gap between vascular translocation (long distance transport) and real-time observation of metabolites within the cell (nano-sensing). Its linkage with in silico metabolism modelling and other approaches will lead to an improved understanding of the regulation of seed filling, as well as a firm foundation for rational metabolic engineering towards higher and fully predictive crop yields.
Wild-type (varieties Barke, Golden Promise and Bowman), seg8 and Risø13 mutant and transgenic Jekyll down-regulated barley plants (Hordeum vulgare L.) were grown under standard greenhouse conditions at 18 °C with 16 h of light and a relative air humidity of 60%. Determination of developmental stages for developing barley seeds and tissue isolations were performed as described (Radchuk et al., 2006).
All NMR experiments were performed on a 17.6 T wide bore (89 mm) superconducting magnet (Bruker BioSpin, Rheinstetten, Germany) equipped with actively shielded imaging gradients (1 mT/m maximum gradient strength). A custom built double-resonant 13C/1H-NMR coil (inner diameter 5 mm) was used for RF pulse transmission and signal reception.
Stable isotope labelling was performed using the 13C feeding procedure after Ekman et al. (2008). Stems were cut 5 cm below the ear and placed in nutrient solution containing ¼ Murashige and Skoog medium, 10 mm glutamine, 10 mm asparagine, 2 mm MES buffer, pH 6.0 and 100 mm UL-13C12-sucrose (Omicron Biochemicals, South Benol, IN).
Direct metabolic13C imaging was performed with the sequence shown in Figure 1b. For 13C SNR improvement by Nuclear Overhauser Effect, a MLEV-pulse scheme was applied for 500 ms on the 1H channel. A 2.5-mm slice was selected with hermite-shaped excitation and refocusing pulses [excitation : pulse duration (tp) = 771 μs, bandwidth (BW) = 7004 Hz; refocusing: tp = 489 μs, BW = 6994 Hz]. The MLEV scheme was also used for broadband 1H signal decoupling during acquisition. The repetition time (TR) of the sequence was 1.5 s, the echo time (TE) 2.7 ms. 2048 spectral points were acquired at a receiver BW of 50 kHz. The experiments were performed with an isotropic FOV of 8 mm using an acquisition-weighted k-space sampling scheme. In the dynamic studies, the number of total scans (NS) was 1200, resulted in an in-plane resolution of 0.42 × 0.42 mm2 and experiment duration (Ttot) of 30 min.
For Inverse metabolic13C imaging, the spatial resolved geHMQC pulse program is shown in Figure 1c. For efficient suppression of the water signal in the 1H spectrum, the sequence starts with frequency selective saturation of the 1H water signal using the VAPOR scheme (Tkác et al., 1999). The first selective 90°1H pulse excites a slice and is followed by the first 90°13C pulse, which creates multiple quantum coherences (MQCs). Spatial phase encoding in the two dimensions was applied after slice selection. The two pulses are separated by τ = 1/(2J). The ratio of the coherence selection gradients G1, G2 and G3 were chosen in such a way (G1: G2:G3 = 2 : 2 : −1) that 50% of the 1H magnetization coupled to 13C spins is refocused. For the spectroscopic imaging geHMQC sequence, two types of experiments were used. For the metabolic uptake (flux) experiments, the TR = 1.5 s, TE = 8.2 ms, τ = 2.9 ms, NS = 800, FOV = 8 × 8 mm2, spatial resolution 0.42 × 0.42 mm2, Ttot = 20 min. For the measurements after 12-h incubation with 13C sucrose, the geHMQC sequence was applied with a higher resolution (0.25 × 0.25 × 1.5 mm3) and a longer experimental time (NS = 7500, Ttot = 3 h 7 min).
Structural1H imaging of the barley seeds was performed by 1H imaging using a multi-slice spin echo sequence. The TR of the experiment was 5 s, TE = 5.1 ms, FOV = 8 × 8 mm2, and 192 × 192 spatial points were acquired, resulting in an in-plane resolution of 40 × 40 μm2. The slice thickness was 200 μm, number of slices = 55, NA = 1, Ttot = 16 min.
For image analysis, in-house software (written in Java) was used for reconstruction and visualization of the aquired NMR data. An exponential time-domain filter (time constant was 1.3 ms for direct 13C measurements and 3.3 ms for indirect 13C measurements) was applied to datasets containing spectral information. To further improve the SNR in the dynamic study, the data for every frame was averaged from seven consecutive experiments. The data from these frames were linear interpolated to create additional frames for a smoother video. Resulting colour-coded images were multiplied with the grayscale reference images to keep the structure visible in the final images. 3D Plots were created using MATLAB 7.9.0 (The MathWorks, Natick, MA). Graphs in Figures 1 and 3 were plotted using Mathematica.
To measure the velocities of sugar allocation, a line along the investigated direction was chosen (Figure 3a). Furthermore, a threshold well above noise level was defined to create an isoline plot. The intersections of the line and the isolines were recorded for all available frames. The velocity was derived from these intersections using linear regression.
Biochemical procedures for isotope analysis
For analysis of 13C labelling using mass spectrometry, seeds were freeze-dried, weighed, grounded and analysed for the content of total carbon and the 12/13C-isotope ratio using elemental analysis (Vario EL3; Elementar Analysesysteme, Hanau, Germany) coupled to isotope ratio mass spectrometry (ESD-100; IPI, Bremen, Germany). The proportion of label that is the result of natural abundance was removed from the data presented. For five seeds, total carbon content and 12/13C-isotope ratio was measured in hand-dissected tissues (pericarp, endosperm, embryo). By relating these data to the tissue weight, the 13C label uptake was calculated for the individual tissues.
In some cases, 13C labelling of starch was measured: pulverized seed material was extracted three-times with ethanol (100%). The remaining starch-containing pellet was dried, solubilized in 1 N KOH for 1 h at 95 °C and incubated with 14 U amyloglucosidase in 1 mL 50 mm sodium acetate (pH 4.8) for 24 h at 55 °C. After centrifugation (10 min, 14 000 g), the supernatant was collected and dried, giving the starch fraction. The pellet was analysed by elemental analysis coupled to ESD-100 for their 12/13C-isotope ratio.
For analysis of 12C/13C labelled sugars, caryopses were homogenized in a Retsch TissueLyser (Qiagen, Hilden, Germany) and extracted with 1 : 1 (v/v) methanol/water. The extracts were filtered with Vivaclear centrifugal filters (Satorius, Göttingen, Germany) at 2000 g for 2 min. Samples were analysed in a LC-MS/MS system (Dionex Ulimate 3000 RSLC; Dionex, Sunnyvale, CA and API 4000; Applied Biosystems, ON, Canada). One microlitre was injected onto a LUNA-NH2 (5 μm, 150 × 2 mm; Phenomenex, Torrance, CA) attached to a precolumn (4 × 2 mm, Phenomenex). As mobile phases acetonitrile (solvent A) and water (solvent B) were used with a flow of 0.4 mL/min in a gradient mode: time/concentration (min/%) for B: 0.0/10; 1.0/10; 2.0/55; 3.5/55; 3.7/10; 5.0/10. The MS was used in negative ion mode, and ions were detected by multiple reaction monitoring. The following transitions were observed: substance/Q1 mass/Q2 mass/de-clustering potential/collision energy for sucrose/341/89/-90/-25 and 13C12-sucrose/353/92/-90/-25. Nitrogen was used as a curtain gas, nebulizer gas, heater gas and collision gas. Ion spray voltage was set to −4200 V, the capillary temperature was 300 °C.
Biochemical procedures for other metabolites, starch and enzyme activities
Extraction and analysis of (unlabelled) metabolic intermediates as well as starch was performed as detailed earlier (Rolletschek et al., 2005). In brief, samples were powdered in liquid N2 and extracted with trichloroacetic acid. Free amino acids were measured by HPLC, soluble sugars by spectrophotometry and metabolites of glycolysis, citric acid cycle and nucleotides by liquid chromatography coupled to mass spectrometry (LC-MS). Starch was determined spectrophotometrically in the pellet remaining after extraction (Heim et al., 1993). The hexose-to-sucrose ratio in the apoplastic liquid was measured by ion chromatography in laser microdissected samples according to Melkus et al. (2009). Extraction and activity assays for enzymes were performed according to previously described protocols: sucrose synthase (E.C. 220.127.116.11; Rolletschek et al., 2005), cell wall-bound and soluble invertase (EC 18.104.22.168; Weschke et al., 2003).
Histochemical techniques and in situ hybridization
Caryopses were fixed, sectioned and stained as described by Radchuk et al. (2006). The electron microscopy was performed as described earlier (Radchuk et al., 2006). Shortly, after staining with 4% uranyl acetate and lead citrate, grids were examined in Tecnai20 electron microscope (FEI, Eindhoven, the Netherlands) at 120 kV. Digital recordings were made with a Megaview III (Soft Imaging Systems, Münster, Germany).
Flux balance analysis
In silico knockout studies were performed based on a recently established FBA model of primary metabolism in the barley seed (Grafahrend-Belau et al., 2009a). To better meet our demands, the model was extended by integrating relevant study specific reactions including ADP glucose, UDP glucose accumulation, cell wall invertase and hexose transporters. Apart from changes based on experimental measurements and literature described in the following, biomass composition and influxes were used as in Grafahrend-Belau et al. (2009a).
All simulations were performed by using the maximization of biomass per flux unit as the objective function (Grafahrend-Belau et al., 2009a) and by constraining the exchange flux for sucrose at representative values for WT (8 μmol/g DW/h; Felker et al., 1984) as well as experimental results. Based on the reduced velocity of sucrose relocation (determined by NMR), the sucrose uptake capacity of caryopsis was set to 3.9 μmol/g DW/h in Jekyll down-regulated plants, and 5.8 μmol/g DW/h in the Risø13 mutant. The upper bound of the uptake rate of the remaining substrates (i.e. Asp, Gln, Ala) was constrained by reducing the simulated WT flux of the respective reaction in correspondence to the experimentally derived reduction of sucrose influx (uptake rates in μmol/g DW/h): WT: Asp: 0.3, Gln: 0.2, Ala: 2.2; Jekyll down-regulated plants: Asp: 0.1, Gln: 0.1, Ala: 1.1; and Risø13 mutant: Asp: 0.2, Gln: 0.1, Ala: 1.6. The upper bound of the O2 uptake rate was fixed at 10 μmol/g DW/h, a seed tissue-specific value derived from experimental results. In silico knockout of the plastidial ADP glucose transporter in the Risø13 mutant was performed by constraining the upper bound of the respective reaction to a 90% reduction of the respective WT flux (Patron et al., 2004). Furthermore, the accumulation of ADP glucose and UDP glucose was fixed, taking the proportion of both metabolites reported in Patron et al. (2004) as a basis. Simulations as well as visualization of the resulting flux maps were performed using FBA-SimVis (Grafahrend-Belau et al., 2009b).
cDNA macroarray and data analysis
Total RNA was extracted from the tissue fractions collected from crease, dorsal pericarp and endosperm fractions from 12 days after anthesis (DAA) using Gentra RNA isolation kit. The isolated RNA was treated with RNAase-free DNase, purified using an Rneasy plant mini kit (Qiagen, Hilden, Germany) and the purified RNA (35 μg) was used for the synthesis of 33P labelled probes. Probe preparation, hybridization and processing of 12 K barley seed array was performed as described in Sreenivasulu et al. (2006).
Images of hybridized nylon membranes were subjected to automatic spot detection using customized MATLAB programs and scored the signal intensities of 11 787 genes from the double spots, enabled us for assessing technical replication. In parallel, we isolated RNA samples from independently grown samples and hybridized to 12 K barley seed array to check biological reproducibility. Quantile normalization was carried out on the complete data set (Bolstad et al., 2003). During further analysis, genes with marginal expression values or diverging double spot ratios were discarded. Fold changes between crease region versus dorsal pericarp region were calculated from two technical and two biological replicates, identified the differentially expressed genes with at least threefold difference and cross-checked for biological reproducibility. Furthermore, the quantile normalized gene expression data obtained from WT and Jekyll transgenic line N91 were analysed from the crease region and identified differentially expressed genes with a fold change >2 and P values were calculated using Benjamini-Hochberg correlation with a threshold of 0.1 to discover the false discovery rate. The quantile normalized values of differentially expressed genes were shown as heat maps in Figure 5g and the detailed data provided in Table S3. Functional annotations were further refined from Sreenivasulu et al. (2006).
We thank U. Wobus for continuous support and discussions. Special thanks to K. Blaschek, U. Siebert and A. Stegmann for excellent technical assistance and to U. Tiemann and K. Lipfert for artwork. We thank F. Fiedler for help with NMR, T. Czauderna for array image processing, and M. Strickert for normalizing the expression data. We acknowledge funding from the Federal Ministry of Education and Research (grants GABI-sysSEED and GABI-GRAIN).