In order to understand sucrose transport in developing seeds of cereals at the molecular level, we cloned from a caryopses library two cDNAs encoding sucrose transporters, designated HvSUT1 and HvSUT2. Sucrose uptake activity was confirmed by heterologous expression in yeast. Both transporter genes are expressed in maternal as well as filial tissues. In a series of in situ hybridizations we analysed the cell type-specific expression in developing seeds. HvSUT1 is preferentially expressed in caryopses in the cells of the nucellar projection and the endospermal transfer layer, which represent the sites of sucrose exchange between the maternal and the filial generation and are characterized by transfer cell formation. HvSUT2 is expressed in all sink and source tissues analysed and may have a general housekeeping role. The rapid induction of HvSUT1 gene expression in caryopses at approximately 5–6 days after fertilization coincides with increasing levels of sucrose as well as sucrose synthase mRNA and activity, and occurs immediately before the onset of rapid starch accumulation within the endosperm. Starch biosynthesis requires sucrose to be imported into the endosperm, as direct precursor for starch synthesis and to promote storage-associated processes. We discuss the possible role of HvSUT1 as a control element for the endospermal sucrose concentration.
Developing seeds synthesize storage compounds from imported sucrose during their maturation phase. Phloem unloading and seed-located transport and transfer processes play an important role during seed development ( Patrick & Offler 1995; Weber et al. 1998a ). Because in seeds the filial generation is apoplastically isolated from the maternal tissue, membrane-located transport steps are necessary. Sucrose transporters have been characterized in a large number of mainly dicot species (for a recent review see Ward et al. 1998 ), and are highly hydrophobic proteins consisting of 12 trans-membrane domains ( Marger & Saier 1993). All sucrose transporters analysed so far are energy-dependent and function as sucrose H+ co-transporters. They represent high-affinity uptake systems with a Km of around 1 m m and are able to accumulate sucrose against a large concentration gradient. Sucrose transporters are encoded by small gene families. They are frequently expressed in the sieve element–companion cell complex, and promote phloem loading. Some members are found to be preferentially expressed in import zones of sink organs where they may catalyse either influx or retrieval of sucrose. However, to date a clear rate-limiting function in sink organs has not been directly proven. Sink-specific sucrose-transport systems were identified in developing seeds ( Gahrtz et al. 1996 ) and anthers ( Stadler et al. 1999 ), and have been characterized in seeds of Vicia faba ( Harrington et al. 1997 ; Weber et al. 1997 ) and pea ( Tegeder et al. 1999 ). Most of the information on sink-specific sucrose transporters comes from dicot seeds. From monocot species, sucrose transporters have been characterized in rice ( Hirose et al. 1997 ) and maize ( Aoki et al. 1999 ).
The cellular pathway of assimilate transport into developing wheat or barley seeds has been analysed in some detail. Photoassimilate exchange is restricted to a single vascular bundle located at the bottom of the crease, extending across the whole length of the grain. Alternative transport pathways are prevented because a cuticle is formed within the cell walls of the integuments surrounding the dorsal region of the endosperm, except at the crease vein area ( Zee & O'Brien 1970). The crease vein is thought to be the site of phloem unloading ( Thorne 1985). Photoassimilates are transported symplastically inwards and are unloaded into the endospermal cavity. The cellular sites of efflux are the membranes of the nucellar projection cells at a position proximal to the endosperm. These cells develop wall ingrowths to amplify the membrane surface ( Wang et al. 1994 ). In wheat, the entire process of sucrose movement through the maternal seed tissue from sieve tube to endospermal cavity appears to be passive and reversible, and does not involve energized membrane transport ( Wang & Fisher 1995). In developing maize kernels, the translocated sucrose is cleaved by invertases during entry into the endosperm and hexoses diffuse throughout the endosperm. Inside the starch-accumulating cells the hexoses are reconverted to sucrose ( Shannon 1972). Unlike in maize, there is no evidence for an invertase-mediated unloading process during the phase of storage product accumulation in barley endosperm ( Thorne 1985). During endosperm development the outermost cell layers differentiate into the aleurone. The aleurone cells at the ventral region adjacent to the crease have thicker cell walls and are termed ‘modified aleurone cells’ ( Cochrane & Duffus 1980). Tracer movement studies suggest that these specialized transfer cells are involved in solute uptake. Inhibitor experiments indicate that sugar uptake into the endosperm is carrier-mediated, probably by an H+ sucrose co-transporter ( Wang et al. 1995 ). Transport is energized by the proton motive force generated by a co-localized H+–ATPase. The modified aleurone cells can therefore function as a complex to accumulate sucrose. On the other hand, the subsequent transfer from the modified aleurone cells to the starchy endosperm is thought to be symplastic ( Wang et al. 1995 ).
As an approach to understanding seed-located sucrose transport in cereals at the molecular level, we cloned two cDNAs encoding sucrose transporters from a barley caryopses library. We analysed the cellular site of expression within developing caryopses, as well as the kinetic properties of their gene products. The spatial and developmental pattern of expression was related to the steady-state levels of sucrose and starch. In addition, the activities of sucrose synthase and cell wall-bound, as well as vacuolar, invertase were monitored. Evidence is presented of an important role of sucrose transporters in sucrose accumulation and the induction of storage activity in barley seeds.
cDNA cloning reveals two sucrose transporter genes expressed in the developing barley caryopses
In order to isolate seed-specific sucrose transporter cDNAs from barley, conserved domains of known sequences from different species were used to design degenerated primers for RT–PCR. Total RNA from developing barley caryopses was used as a template for PCR. Amplified cDNA fragments of around 200 bp were subcloned and sequenced. One of the cloned fragments was homologous to sucrose transporter sequences and was used as a probe to screen a cDNA library generated from developing caryopses. Hybridization under low-stringency conditions revealed 29 independent positive clones which contained a single type of a Hordeum vulgare putative sucrose transporter cDNA (HvSUT1). Only one clone, identified by only a very faint hybridization signal, was clearly different from HvSUT1, therefore representing a putative second sucrose transporter cDNA. The 3′ non-translated region of this clone was used to screen the library again under high-stringency conditions. Ten independent positive clones were isolated, representing a single mRNA species. We designated this cDNA sequence as HvSUT2. The HvSUT1 and HvSUT2 cDNAs possess a single open reading frame of 1694 and 1521 bp encoding 523 and 507 AS, respectively. The cDNA organization is shown in Fig. 1(a). The putative amino-acid sequences of both HvSUT1 and HvSUT2 were aligned with the deduced protein sequences of 15 known sucrose transporters ( Fig. 1b). HvSUT1 was highly homologous (81.3 and 82.9% identity) to OsSUT1 from rice ( Hirose et al. 1997 ) and to ZmSUT1 from maize ( Aoki et al. 1999 ). However, HvSUT1 and HvSUT2 show only 42.2% similarity. HvSUT2 did not cluster with the monocot transporters HvSUT1, ZmSUT1 and OsSUT1, nor with the dicot transporter family, but appears to form an independent group.
HvSUT1 and HvSUT2 mediate sucrose transport into yeast cells
The cDNAs encoding HvSUT1 and HvSUT2 were expressed in baker's yeast in both sense and antisense orientations. The yeast strains HVY-S1 and HVY-S2, which expressed HvSUT1 and HvSUT2 in sense orientation, were able to transport sucrose across the yeast plasma membrane, whereas only a very slow uptake rate was seen in the HVY-A antisense strains ( Fig. 2a,b). The Km values for HvSUT1 and HvSUT2 were 7.5 and 5 m m, respectively, somewhat higher compared to the transporters from dicot species. Sucrose uptake into both HVY-S1 and HVY-S2 yeast strains was inhibited by p-chloromercuribenzoylsulphonic acid and carbonyl cyanide-m-chlorophenylhydrazone (data not shown). The uptake experiments indicate that both HvSUT1 and HvSUT2 encode functional sucrose transporters.
Expression analysis of HvSUT1 and HvSUT2
Sucrose transporters in higher plants are encoded by small gene families. We therefore used the 3′ non-translated regions of both HvSUT1 and HvSUT2 as gene-specific probes in RNA gel-blot experiments. A single band of 2.3 and 2.0 kb was detected with the HvSUT1- and HvSUT2-specific probes, respectively. Transcripts from both genes were found in sink and source organs. However, HvSUT1 mRNA levels were very high in caryopses 7–11 days after flowering (DAF) and much lower in sink and source leaves and in roots ( Fig. 3a). The level of HvSUT2 mRNA was highest in developing leaves, but the transcript was also present in comparable amounts in all other plant organs analysed ( Fig. 3b). Compared to HvSUT1, the amount of HvSUT2 mRNA was relatively higher in younger (0–6 DAF) and older (18–21 DAF) caryopses. Remarkably, in the developing caryopses mRNA levels of both transporters were transiently lower at 9 DAF. This was confirmed by three independent experiments. Taken together, the Northern data indicate that HvSUT1 is predominantly expressed in developing caryopses during mid-development. In contrast, HvSUT2 is expressed to nearly equal levels in all tissues analysed.
Histological localization of HvSUT1 and HvSUT2 mRNAs
Figure 4 shows a transverse section through the middle region of a caryopsis at 6 DAF. The well developed pericarp (P) represents the main part of the caryopsis. The embryo sac is surrounded by three different tissue layers, the outer (OI) and inner integument (II), each composed of two cell layers and the mono-layered nucellar epidermis (NE). At this developmental stage the endosperm has already differentiated into the starchy endosperm (SE) and the endospermal transfer cells (ET). In front of the endospermal transfer cells, the nucellar epidermis is changed into the so-called nucellar projections (NP). Further conspicuous elements of the grain are the large main vascular bundle, and three smaller veins, two at the lateral sides and one at the dorsal side (VT). From 7 DAF the pericarp becomes progressively degraded in its inner region, contacting the outer integument and later also including the lateral and dorsal veins. In parallel, filling of the starchy endosperm begins. Both processes lead to a shift in the relative volume of pericarp and endosperm in favour of the latter and coincide with a complete metabolization of the starch of the pericarp. At the end of development nearly the whole caryopsis consists of starchy endosperm.
Cell type-specific expression was analysed in developing caryopses by in situ hybridization using the 3′ non-translated regions as probes. The distal part of a longitudinal section of a grain at 6 DAF shows the micropylar region ( Fig. 5a). Transverse sections through the middle region of a grain at the same developmental stage show the pericarp ( Fig. 5d), the outer and inner integument and the mono-layered nucellar epidermis ( Fig. 5j). The endospermal transfer cells with the underlying starchy endosperm directly face the nucellar projection cells ( Fig. 5d,g). HvSUT1 mRNA accumulation was highest in the cells of the endospermal transfer layer and of the nucellar projections ( Fig. 5e,h). In addition, weak HvSUT1-specific label was present in cells of the inner integument and the nucellar epidermis ( Fig. 5k), whereas the pericarp was occasionally found to be slightly labelled. The overall distribution of the HvSUT2-specific label was spatially similar to HvSUT1 except that cells of the pericarp and of the endosperm transfer layer were marked to nearly equal intensities ( Fig. 5f,i). In contrast to HvSUT1, the HvSUT2-specific labelling was very abundant in the micropylar region of the pericarp ( Fig. 5b,c). In addition, there was also weak labelling of the nucellar epidermis, especially in those cells flanking the nucellar projections on both sides ( Fig. 5i). For both transporters, labelling was completely absent from cells of the vascular system of the caryopsis. The in situ hybridization data indicate that HvSUT1, compared with HvSUT2, displays a much more specific expression pattern within developing caryopses. HvSUT1 is mainly expressed in the nucellar projections and in the endospermal transfer cells which represent the putative sites of membrane exchange of assimilates between the maternal and the filial generations.
An electron microscopic examination of those tissues which accumulate HvSUT1 and HvSUT2 mRNA provided ultrastructural indications for active transport processes such as cell-wall ingrowths and wall foldings. Cells of the nucellar projection exhibiting thick walls with strong, irregularly shaped invaginations form a narrow zone between a layer of desintegrated cells at the dorsal side bordering the endospermal cavity, and the ventrally situated thin-walled cells ( Fig. 6a,b), which are connected via numerous plasmodesmata (data not shown). The endospermal transfer cells can easily be distinguished from the adjacent cells of the starchy endosperm because they possess a dense cytoplasm and thick walls, with only a few ingrowths ( Fig. 6a). The cells of the inner integument possess characteristically very thin and multiple-folded anticlinal walls, resulting in an up to sevenfold increase of their length ( Fig. 6c,d). Wall folding was restricted to the longer cells at both sides of the nucellar projection which accumulate transporter mRNA. Transfer cell morphology could not be found in the pericarp (data not shown).
The time-course of endosperm-specific starch accumulation correlates with sucrose levels
Fresh weight increase in developing caryopses was measured at 1-day intervals from flowering until desiccation ( Fig. 7a). Around 10 DAF, a 3-day lag phase became evident only when seed development was retarded by growing the plants at lower temperature ( Fig. 7a). As shown for the seed development of legumes, these lag phases often mark switch points in seed development ( Hedley & Ambrose 1980).
The accumulation pattern of starch in caryopses was studied by iodine staining of hand-cut sections. Starch was already present prior to anthesis, and was found in medial–longitudinal sections in the micropylar part of the pericarp except in tissue surrounding the micropylar channel ( Fig. 8a,g). For the following stages, transverse sections from the mid-part of the grain were used. At 2 DAF starch was concentrated around both lateral vascular bundles ( Fig. 8b). From 3–4 DAF the tissue surrounding the dorsal vascular bundle also became involved in starch accumulation ( Fig. 8c,d,h). At 5–6 DAF the maximum accumulation was observed in the wings ( Fig. 8e,f,h,i). At this time the endosperm which fills the whole embryo sac cavity begins accumulation. Starch grains appeared first in the centre of the endosperm, within the cells that are last to form a wall ( Olsen et al. 1992 ; Fig. 8f,i), spreading at 7 DAF throughout the whole starchy endosperm. In the pericarp the level of starch dropped and could hardly be detected at 8 DAF (data not shown).
We measured glucose, sucrose and starch within developing caryopses. Glucose levels in caryopses were high at earlier stages (3–10 DAF) and decreased when the starchy endosperm entered the starch filling phase. The levels were particularly low in the embryo sac at 3 DAF, and again at 11 DAF onwards during the linear phase of starch accumulation ( Fig. 7b). Sucrose in both caryopses and embryo sac increased from 3 DAF and reached a maximum at 7 DAF ( Fig. 7c). This pattern roughly followed the accumulation profile of HvSUT1-mRNA (grey bars, Fig. 7c). After 7 DAF sucrose dropped in the caryopsis and was less pronounced in the embyo sac, but remained at constant high levels of 40–60 μmol g−1. After 6–7 DAF starch increased continuously in the caryopses ( Fig. 7d), which could be attributed mainly to the endosperm because levels in the pericarp were very low between 7 and 12 DAF ( Fig. 7d).
Sucrose synthase but not invertase is associated with starch accumulation
We analysed the enzyme activity of the sucrose-cleaving enzyme sucrose synthase (SUS), known to be one of the major enzymes in the sucrose-to-starch pathway and cell wall-bound invertase as well as vacuolar invertase. The activity of vacuolar invertase was found to be high very early in development (2 DAF), then dropped rapidly and remained at a low level during the mid-to-late phase of caryopsis development (6–21 DAF) ( Fig. 9a). As shown for different seed tissues ( Fig. 9a, inset), most of the measured activity of the vacuolar invertase was located within the pericarp. The enzyme activity of vacuolar invertase determined in the starchy endosperm and in the embryo sac, as well as in the whole caryopsis at later developmental stages, was found to be near the detection limit. Enzyme activity of cell wall-bound invertase was present throughout the whole development in caryopses, at decreasing levels ( Fig. 9b). The analysis of different seed tissues revealed that most of the activity between 3 and 13 DAF was attributed to the pericarp. In the embryo sac the relatively high activity level at 7 DAF decreased to very low values at 12–13 DAF, while in dissected starchy endosperm, cell wall-bound invertase activity at 7–13 DAF was near the detection limit ( Fig. 9b, inset).
The enzyme activity of SUS was very low in young caryopses, but increased dramatically by about a factor of 10–15 between 5 and 9 DAF and remained high until 21 DAF ( Fig. 9c). SUS activity could be attributed mainly to the endosperm because of low activity in the pericarp ( Fig. 9c, inset).
Taken together, the data indicate that in developing caryopses the highest amount of glucose is present before starch synthesis is initiated, whereas during starch synthesis sucrose is the predominant sugar. The activity of both invertase isoforms is confined mainly to the pericarp but is very low or absent within the starchy endosperm. SUS activity, on the other hand, is associated mainly with the endosperm and correlates with the starch accumulation phase. Moreover, the rapid induction of SUS enzyme activity is correlated to HvSUT1 gene expression and high sucrose levels.
During seed development of cereals the endosperm becomes a highly differentiated starch-storage organ. To accomplish these biosynthetic functions it is necessary to import sucrose as precursor. In this paper we provide evidence that seed-located sucrose transporters are important for this process in barley.
HvSUT1 is preferentially expressed in caryopses
A PCR-based cloning approach and a subsequent screening of a barley caryopses-specific cDNA library revealed two different cDNAs encoding sucrose transporters, HvSUT1 and HvSUT2. The alignment of their primary sequences shows that HvSUT1 is closely related to OsSUT1 from rice ( Hirose et al. 1997 ) and to ZmSUT1 from maize ( Aoki et al. 1999 ). Interestingly, HvSUT2 fits into neither the monocot (OsSUT1, ZmSUT1 and HvSUT1) nor the dicot transporter family, and appears to represent an independent group.
Functional expression in yeast cells and sensitivity against protonophores indicate that both HvSUT1 and HvSUT2 can mediate sucrose uptake and thus probably represent sucrose H+ co-transporters. Compared with other carriers' Km values (1 m m), HvSUT1 and HvSUT2 have a clearly higher Km of 7.5 and 5 m m, respectively. However, because there are no kinetic data available from sucrose carriers of monocot species, these differences could be simply ascribed to specific properties of transporters from monocots.
HvSUT1 gene expression in seeds is confined to cells at the maternal–filial boundary having transfer cell morphology
A detailed expression analysis by RNA gel-blot analysis and by in situ hybridization was done using the 3′ non-translated regions of HvSUT1 and HvSUT2. Therefore any cross-hybridization to a possible third, but so far unidentified, sucrose transporter in barley should be excluded.
The organ-specific expression pattern is strikingly different between HvSUT1 and HvSUT2. HvSUT1 is mainly expressed in developing caryopses but is otherwise not restricted to sink tissues. Moreover, in leaves the expression increases during the sink-to-source transition ( Fig. 3a; W. Weschke, unpublished results). In caryopses HvSUT1 is expressed in the endospermal transfer-cell layer, the nucellar projections bordering the endospermal cavity and at very low levels in the nucellar epidermis and the inner integument. These tissues together represent the maternal–filial boundary, and the two former are thought to be the cellular site where sucrose transport is believed to occur. In addition, endospermal transfer cells and the nucellar projections are characterized by the formation of transfer cell morphology. Transfer cells are formed in the endosperm immediately before the endosperm cells begin starch accumulation ( Huber & Grabe 1987). Around 6 DAF the HvSUT1 gene expression increases dramatically. We therefore speculate that the induction of HvSUT1 gene expression in the barley endosperm is somehow coupled to transfer-cell formation. This is in analogy to developing seeds of V. faba where we have shown that VfSUT1 is expressed during transfer-cell formation in the maternal thin-walled parenchyma of the seed coat and in the adjacent outer cotyledonary epidermis ( Weber et al. 1997 ). We conclude that in developing barley caryopses HvSUT1 is expressed in the cells of the maternal–filial boundary which frequently show transfer-cell morphology and are probably involved in sucrose transfer. This points to an important role of HvSUT1 in controlling sucrose unloading from the maternal tissues and/or loading into the endosperm. In contrast, the expression of the rice OsSUT1 ( Hirose et al. 1997 ) and the maize ZmSUT1 ( Aoki et al. 1999 ) is more confined to the photosynthetically active tissues. These isoforms may therefore be involved in sucrose loading into the phloem in source tissues.
The second transporter gene, HvSUT2, is principally expressed in the same cell types as HvSUT1; but in contrast to HvSUT1, in the caryopses expression occurs not mainly in the cells of the maternal–filial interphase but is, for example, very high in the pericarp. Again, expression is not strictly caryopsis-specific but occurs also in all other sink and source organs analysed. Remarkably, in all tissues analysed HvSUT2 mRNA accumulates to similar levels and could therefore be regarded as constitutively expressed. This points to a more general housekeeping role of HvSUT2.
Unlike the monosaccharide transporters, which are expressed in a very specific manner ( Weig et al. 1994 ), only one sucrose transporter has been identified which is expressed specifically in the sink tissues anthers and gynoecia ( Stadler et al. 1999 ). The others are found in sink as well as source tissues. For example, the ‘sink-specific’ DcSUT2 is expressed predominantly in carrot tap roots but also in leaves ( Shakya & Sturm 1998). The same is true for the ‘cotyledon-specific’ transporter VfSUT1 of faba bean ( Weber et al. 1997 ), for RcSUT1 of Ricinus ( Bick et al. 1998 ; Weig & Komor 1996), and for HvSUT1 and HvSUT2 of barley described here.
HvSUT1 gene expression is correlated to sucrose levels in barley caryopses
It has been shown by physiological experiments that, in developing wheat seeds, the layer of endosperm transfer cells is the principal cellular site of sucrose uptake, and sucrose is actively accumulated across their membranes by a sucrose-H+ symporter ( Wang et al. 1995 ). We show here that in barley caryopses the HvSUT1 gene is expressed in that endospermal transfer cell layer. In addition, gene expression of HvSUT1 increases dramatically between 3 and 7 DAF, by 30–50-fold when transfer-cell formation occurred. At the same time a parallel increase of sucrose concentration was measured. Moreover, between 6 and 15 DAF, the HvSUT1 mRNA levels correspond tightly to the sucrose levels ( Fig. 7b). We therefore conclude that HvSUT1 is predominantly responsible for the accumulation of sucrose by active uptake across the membranes of the endospermal transfer cells. It has been shown for wheat caryopses that the cytochemical localization of H+–ATPase activity is confined to the endospermal transfer cells and therefore could provide the proton-motive force to energize sucrose uptake through these cells ( Baker et al. 1991 ).
Similarly to VfSUT1 of faba bean cotyledons, HvSUT1 is also expressed in cells of the maternal part of the filial–maternal boundary, in the transfer cells of the nucellar projections, and to a lesser level in the nucellar epidermis and the inner integument. However, a specific role of these transporters in sucrose unloading from the maternal tissue into the endospermal cavity is uncertain. Firstly, a simultaneous release of sucrose from maternal tissue and uptake into filial tissues, both mediated by a sucrose–H+ symporter, would be thermodynamically difficult to explain. Secondly, it has been shown, at least for wheat, that the entire process of sucrose unloading from the maternal seed tissues is passive and not mediated by energy-dependent transport ( Wang & Fisher 1995). Therefore the most likely suggestion is that the HvSUT1 in the cells of the nucellar projections is involved in sucrose retrieval back into the maternal tissue. The direction of net sucrose transport would then be dependent on the establishment and alignment of a proton gradient to energize sucrose transport.
Control elements for the endospermal sucrose concentration
In early caryopsis development (2–5 DAF), high levels of hexoses were measured in the whole caryopsis. Because in this developmental period the hexose concentration in the embryo sac is very low, the high hexose concentration measured in the whole caryopsis must be attributed to the pericarp. The high enzymatic activity of both vacuolar and cell wall-bound invertase, located in this developmental period specificially in the pericarp, supports this assumption. A reduction in the level of cell wall-bound invertase was seen at around 5 DAF ( Fig. 9b). In parallel, a relatively high concentration of sucrose was measured in the pericarp ( Fig. 7b), and the pericarp transiently accumulates starch ( Fig. 8). Between 6 and 8 DAF, an approximately 10-fold increase in the hexose level was measured in the embryo sac. This increase in hexose concentration corresponds to the second hexose peak measured in whole caryopses. The hexose concentration peaks around 7–8 DAF, in exactly that developmental phase where an activity maximum of the cell wall-bound invertase in both the pericarp and the embryo sac was detected ( Fig. 9b, inset). A parallel increase in sucrose concentration was measured in the embryo sac, and both sugar profiles are nearly identical up to 8 DAF. However, whereas the hexose level decreases rapidly at 11 DAF, the sucrose level remains high between 40 and 60 μmol g−1 FW during the whole caryopsis development. This causes a change in the hexose : sucrose ratio in favour of the latter, described as characteristic for the induction of the storage phase in V. faba seeds ( Weber et al. 1995 , Weber et al. 1996 ). However, whereas the accumulation process in the starchy endosperm starts around 7 DAF, the increase in the sucrose : hexose ratio is evident only at 11 DAF. Possibly this discrepancy can be explained by specificities of the monocot species barley.
In early caryopsis development, the pericarp transiently accumulates starch ( Fig. 8) and the expression of HvSUT2 in this tissue could be related to this. Similarly in maize, starch is a major constituent of the ovary before the endosperm is filled. Ovary starch pools can play a role in maintaining maize kernel development. When photosynthesis is inhibited, which diminishes sucrose transfer to the ovaries, these starch pools disappear whereupon embryo development is arrested ( Zinselmeier et al. 1999 ).
Later the endosperm becomes the main starch storage organ. In barley, the phase of endospermal starch accumulation starts at 8–9 DAF and is finished at around 17–18 DAF. Endosperm filling is therefore accomplished within little more than 1 week. This rapid starch biosynthesis rate requires large amounts of sugars to be imported into the endosperm. Sucrose is the major sugar in caryopsis and embryo sac during the seed-filling phase, whereas glucose appears to have a less important role because its levels are fairly low ( Fig. 7b,c). The particular role of sucrose is firstly as the direct precursor for starch synthesis in developing seeds and other starch-storing organs such as potato tubers ( Smith et al. 1997 ), and secondly to induce genes of the sucrose-to-starch pathway in sink organs ( Weber et al. 1997 ; Weber et al. 1998b ). We show here that SUS activity in developing barley caryopses is mainly associated with the endosperm. Moreover, the SUS activity level is correlated with the sucrose concentration. On the other hand, the activity of the cell wall-bound invertase is very low in the starchy endosperm, in contrast to the pericarp ( Fig. 9b, inset). In developing maize kernels the activity of a cell wall-bound invertase increases dramatically between 0 and 12 days after pollination, whereas a soluble isoform is not present ( Carlson & Chourey 1999). The tissue- and temporal-specific expression of the cell wall-bound invertase is essential for normal kernel development ( Miller & Chourey 1992). However, endospermal starch accumulation in maize is dependent on SUS. In developing maize endosperm two SUS isoforms are present. SS1 plays the dominant role in providing the substrate for cellulose biosynthesis, whereas the SS2 protein is needed mainly for generating precursors for starch biosynthesis ( Chourey et al. 1998 ).
In faba bean cotyledons accumulation of sucrose synthase mRNA parallels that of sucrose, and its mRNA can be induced in vitro after feeding this sugar ( Heim et al. 1993 ). Decreasing sucrose in developing cotyledons of narbon beans after expressing an alien invertase decreases starch content and down-regulates genes encoding enzymes of the sucrose-to-starch pathway ( Weber et al. 1998b ). Accordingly, for cereals it has been reported that in wheat endosperm the rate of storage starch accumulation is a function of the sucrose concentration, which is tightly controlled ( Jenner et al. 1991 ). It is therefore necessary for the induction of storage-starch biosynthesis to increase sucrose levels. This could be performed by a high-affinity uptake system which is able to accumulate sucrose. We show here that the rapid induction of HvSUT1 gene expression at around 5–6 DAF coincides with increasing sucrose levels. These events take place immediately before the onset of rapid starch accumulation. It is most likely that HvSUT1, as a high-affinity uptake system, can act as a control element for endospermal sucrose concentration.
We present here correlative evidence for an important role of sucrose transporters in sucrose accumulation and starch biosynthesis in barley endosperm. Direct proof of a rate-limiting role for HvSUT1 or any other seed-located sucrose transporter has not been achieved at the moment. This could be best done by antisense expression of the HvSUT genes under control of their own promoter.
Plants from a two-rowed spring barley (Hordeum vulgare cv. Barke) were cultivated in growth chambers. During the generative phase of development the plants were grown under a 16 h light, 20°C and 8 h dark, 14°C regime or, to retard development, at 18 and 10°C during the light and dark phases, respectively. Days after flowering (DAF) were defined by determining anthesis on spikelets in the centre one-third of the spikes. Only five kernels from each row corresponding to this region were used in all the studies presented.
RT–PCR mediated cloning of sucrose transporter cDNAs
The following primers were deduced from conserved regions of dicot sucrose transporter cDNAs (EMBL data library) and one rice EST: primer 1: 5′-TTCTC/TGGIA/GTCCCICTTGC-3′ primer 2: 5′-A/GAAC/TGCC/TGGIAA/TGTTICC-3′. Five μg total RNA from barley caryopses of 8–11 DAF were reverse transcribed for 60 min at 42°C using MMLV reverse transcriptase and oligo (dT)16 primer. The synthesized first-strand cDNA was used for hot-start PCR amplification with 1.5 m m Mg(OAc)2 and 0.4 μM of primers 1 and 2. The temperature regime was 2 min at 95°C for the hot start, 0.5 min at 94°C, 0.5 min at 48°C, 1.5 min at 72°C for 35 cycles, and for final extension 5 min at 72°C. Amplified DNA bands of ∼200 bp were subcloned into pUC18. A cDNA library from barley caryopses at 1–15 DAF was constructed using the ZAP Express cDNA Synthesis Kit (Stratagene) and screened under low-stringency conditions using the PCR-derived cDNA fragment. The cDNA sequences of HvSUT1 and HvSUT2 have EMBL/GenBank/DDBJ accession numbers AJ272309and AJ272308, respectively.
RNA extraction and hybridization procedures
Total RNA was isolated from anthers and the female part of the flower immediately before anthesis, as well as from developing caryopses, growing and mature leaves and roots, as described by Heim et al. (1993) . RNA (15 μg lane−1) was separated on an agarose gel and blotted. Membranes were sequentially hybridized according to Church & Gilbert (1984) using the following probes: two cDNA fragments 384 bp (nucleotides 1634–2017) and 330 bp (nucleotides 1717–2046) in length, representing the 3′ NTR of HvSUT1 and HvSUT2, respectively, and additionally a 26S rDNA fragment to estimate the relative amounts of RNA bound in each lane (data not shown). The hybridization signals were quantified as described by Weber et al. (1996) and are given in relative units. In situ hybridization was performed according to Panitz et al. (1995) using 33P-labelled cDNA fragments specific for the 3′ NTR of HvSUT1 and HvSUT2. For negative control, slides were treated with RNase before hybridization. No labelling of any cell type by any probe was seen (data not shown).
Functional expression of HvSUT1 and HvSUT2 in Saccharomyces cerevisiae
The plasmids NEV-S-HvSUT1, NEV-S-HvSUT2, NEV-A-HvSUT1 and NEV-A-HvSUT2 were transformed into the yeast strain DBY2617 ( Kaiser & Botstein 1986) as described ( Gietz et al. 1992 ), yielding the strains HVY-S1 and HVY-S2 (harbouring the sense constructs) or HVY-A1 and HVY-A2 (harbouring the antisense constructs). Sucrose transport was analysed in 50 m m sodium phosphate buffer as described by Sauer et al. (1990) .
Determination of sugars, and starch and enzyme assays
Procedures were performed as described by Heim et al. (1993) . Briefly, soluble carbohydrates were extracted in 80% ethanol at 80°C and the concentration determined enzymatically. The remaining insoluble material was used for starch determination after dissolving in 1 m KOH and hydrolysing with amyloglucosidase. Cell wall-bound invertase was determined as described by Weber et al. (1996) . Sucrose synthase was measured by monitoring the synthesis of UDP-glucose as described by Weber et al. (1998b) .
Histological methods and electron microscopy
Grains were removed from the mid-region of the ear at 1-day intervals after flowering, the palaea peeled off, and median transverse series (0.5–1.0 mm thick) cut by hand with a razor blade from each caryopsis. Unfixed sections were immersed in acetocarmine (0.5% in 45% acetic acid) for 30–60 sec, washed with distilled water, and stained for starch with iodine–potassium iodine. Microscopic analysis was done using a ZEISS Axioskop with dark-field illumination. For electron microscopy, transverse sections of the middle part of the caryopsis were immersed in fixative consisting of 2% glutaraldehyde in 50 m m cacodylate buffer pH 7.2 for 2.5 h. After washing (four times, 15 min) with cacodylate buffer the tissue was post-fixed in 2% OsO4 in 50 m m cacodylate buffer and in water, 15 min each, and then stained overnight in 2% aqueous solution of uranyl acetate on ice. The specimens were washed three times for 10 min each with double-distilled water, dehydrated in an acetone series, infiltrated with Spurr's resin, polymerized with uranyl acetate and lead citrate, and examined in a Zeiss LEM902A operating at 80 kV.
We are grateful to Hardy Rolletschek for his help in biochemical analysis. We thank Angela Stegmann, Elsa Fessel, Katrin Blaschek and Carola Hümmer for excellent technical assistance. This work was supported by the Land Sachsen-Anhalt (FKZ 2678/A/0087G). U.W. acknowledges additional support by the Fonds der Chemischen Industrie.