Introduction

Wheat grain is a single-seeded fruit in which the major tissue is the starchy endosperm, accounting for approximately 83–84% of the dry weight. In contrast, the embryo, the aleurone, and the outer layers (pericarp and testa) account for approximately 3%, 6.5%, and 8% of the dry weight, respectively1. The starch content accounts for 60–70% of wheat grains, providing approximately 25% of the calories for humankind (http://www.fao.org/faostat/zh/#data/FBS). Additionally, the content and composition of starch are crucial for the baking quality of flour. Thus, modification of the starch content and composition of wheat presents an opportunity for potentially large-scale improvements in flour quality and end-use.

Interestingly, although the weight of a wheat grain is only approximately 40 mg, the distribution of nutrients in the grain is not homogeneous. The grains exhibit compositional gradients from the outer to the inner parts, as confirmed by various imaging approaches and biochemical analyses of the pearling fractions. For example, the starch content increases with changes in the proportions of amylose and amylopectin from the outer to the inner fractions2. The two types of starch granules are also unevenly distributed across the endosperm, with subaleurone cells containing greater proportions of B-type granules (<10 µm) than central starchy endosperm cells3. The differences in starch composition and starch granule size in the pearling fractions are expected to influence the functional properties of flour. In regard to baking, the inner part of the grain is most suitable for biscuit making due to its high starch content and low gluten content4. However, until recently, studies were limited to the gross distribution of starch content, composition, and processing quality.

Starch biosynthesis in reserve organs is characterized by a high rate of starch production from sucrose and the efficient formation of starch granules, including amylopectin and amylose5. Thus, it is reasonable to hypothesize that the supply of sucrose could restrict starch biosynthesis in different parts of the grain. It has been reported that the decrease in sucrose concentration from the endosperm cavity to the dorsal surface is much steeper than that of starch6. The authors concluded that the pattern of starch deposition throughout the wheat endosperm could not be ascribed to a regional pattern of sucrose concentration. However, this work did not explore the spatial distribution of sugars before 14 DAA, when the starch content started to increase rapidly7. Hence, it is necessary to trace the changes in sugars in different parts of the endosperm during the filling stage, especially at the early stage before starch synthesis initiates.

Based on the importance of sucrose for starch gradient formation, it is crucial to explore how sucrose is transported from the mother plant into the endosperm. Substrates, including sucrose from mother plants, unload at the cavity first, followed by translocation into the endosperm. Two transport pathways for substrate entering the endosperm have been suggested (Supplementary Fig. 1): (1) radial transport across the tissue from the cavity and cell transfer and/or (2) transport via aleurone around the endosperm8. It remains controversial whether these two pathways transport different nutrients9,10,11. Nondestructive imaging tools that observe substrate transport and allocation in grains on appropriate spatial and temporal scales facilitate investigations of transport pathways. Recently, positron emission tomography-computed tomography (PET-CT) has been developed as a practical imaging technique for assessing entire plant transport and nutrient allocation by quantifying the distribution of positron-emitting radioisotopes in a plant noninvasively and over time12,13,14. PET is a key tool needed to identify radioisotopes such as carbon-11(11C) and nitrogen-13 (13N) with a spatial resolution of 1.4 mm, and CT provides precise spatial orientation with a spatial resolution of 20 µm. Currently, PET-CT has been applied to plant studies, focusing on the transport and allocation of photoassimilates in sorghum15, giant reed16, and tobacco17. However, PET-CT is rarely used in grain imaging studies to trace sugar dynamics.

Another hypothesis is that the differences in the conversion of sugars into starch in different layers of the wheat grain results in spatial gradients of starch. Starch biosynthesis is carried out by the orchestrated action of enzymes, including starch synthases (SSs), starch branching enzymes (SBEs), and debranching starch enzymes (DBEs)18. However, information on the spatial expression profiles of genes encoding genes involved in starch biosynthesis is still limited due to the poor accessibility of transcripts from different plant parts within developing caryopses. Recently, the combination of laser microdissection and transcriptome analysis has been used to monitor spatial gene expression profiles in different plant tissues, such as transfer cells and nucellar cells of endosperm in barley19, cells of the lateral root in maize20 and central cells of embryo in Arabidopsis21. With these techniques, it is possible to obtain more detailed insight into the spatial differences in starch and sugar metabolism in developing wheat caryopses.

This research was undertaken to classify tissue-specific variations in metabolites and transcripts involved in starch biosynthesis to investigate the factors affecting the spatial distribution of starch in wheat grain. The transport pathway of sugar in developing wheat spikes was monitored nondestructively via micro PET-CT to provide the first description of the dynamics of sugar transport in relation to starch deposition. The aims of this study were to investigate (1) sugar substrate supply patterns and transport pathways during the filling stage and (2) starch synthesis capacity in different wheat endosperm layers. The results provide important insights into potential mechanisms that affect starch metabolism and starch content in wheat grain tissues, especially from a spatial perspective.

Results

Spatial gradients of starch and starch components exist inwardly along the wheat endosperm

The contents of total starch and starch components showed large variations within different parts of the grains (Table 1). Overall, the contents of total starch and amylopectin increased inwardly, while no significant difference was observed within the endosperm in terms of amylose content. The divergence in starch content resulted in different RVA parameters. Compared with those from the endosperm, the flour from the aleurone group showed distinct pasting characteristics, including lower viscosities, breakdown, and extremely higher pasting temperatures. The flour from the inner endosperm region possessed a greater trough and final viscosity, greater setback, and lower breakdown than did the flour from the outer endosperm.

Table 1 Spatial distribution of starch and starch components in mature wheat grains
Full size table

PET-CT-monitored transportation of glucose from the endosperm cavity to the caryopsis reveals the importance of the starch biosynthesis substrate supply in the spatial gradient of starch in the endosperm

The starting metabolite for starch synthesis in amyloplasts is sucrose, which is translocated to the reserve tissues from leaves through the phloem5. To monitor the transportation of sucrose in caryopses instantaneously, we used in vitro spike cultivation combined with a commercial micro PET-CT scanner. Since it is currently difficult to label sucrose with positron-emitting radioisotopes (such as 11C or 13N), a common radioisotope, [18F]FDG, was used instead.

To demonstrate that glucose can be absorbed by the production of caryopses and used for starch synthesis, in vitro glucose spike cultivation was first performed. The spikes were detached immediately after the heading stage. Glucose or sucrose has been applied to media as a single sugar source. The caryopses cultivated in glucose or sucrose media developed normally at 7 DAA and 15 DAA (Supplementary Fig. 2). The starch content in mature grains cultivated in glucose media was slightly lower than that in mature grains cultivated in sucrose media, indicating that glucose could be used to simulate sucrose transport in wheat caryopses.

The spikes were detached from plants at 9, 15, and 25 DAA. To eliminate the contribution of sucrose from leaf photosynthesis to achieve more careful control of the carbon supply, the stems were cut to approximately 15 cm (spikes not included), and all the leaves were excluded from the first node. The cut stems were placed into a 50 ml centrifuge tube filled with media without a sugar supply. [18F]FDG (300 microcuries) was injected into the tube immediately, after which the tube was placed into a micro-PET-CT system (Fig. 1a). We acquired dynamic images of the panicles for 90 min (in 1 min bins) (for movies, see Supplementary Movie 1).

Fig. 1: Experimental setup for developing wheat spikes (15 DAA) for PET imaging using a commercial micro PET-CT scanner.
figure 1

a In vitro spike cultivation and side view of spikes set up in micro PET-CT. b CT image of the side view and vertical view. c A reconstructed PET image of the 18F-FDG distribution over time in a wheat spike. Scale bar = 3 mm.

Full size image

The transection of panicles was acquired by CT, as shown in Fig. 1b, where we could clearly see the rachis, cavity, and sections of the caryopses. The radioactivity acquired by PET is shown as a color that overlaps with that of CT (Fig. 1c). After absorbing media for 30 min, high radioactivity was apparent at the rachis, and weak radioactivity appeared at the rachilla, indicating that [18F]FDG was transported into the rachis, followed by the rachilla. At 65 min, the radioactivity in the endosperm became stronger but was only distributed around the cavity. Afterward, the [18F]FDG was transported into the whole endosperm, where the radioactivity was highest in the inner endosperm, followed by the middle endosperm, and finally the outer endosperm. At the same time, it was almost invisible in the aleurone and seed coats. To rule out the possibility that [18F]FDG was transported underwater flow, we also investigated the transport pathway of [18 F]FPGLN22, which showed a completely different pattern of [18F]FDG (Supplementary Fig. 3). All the results confirmed that after downloading into the cavity, [18F]FDG was transported along the inner, middle, and outer endosperm and finally into the aleurone.

The expression of sugar transporter-encoding genes and sugar contents in TC and AL further confirmed the importance of substrate supply in the spatial gradient of starch in the endosperm

To support the transport pathway observed by PET-CT, we further compared the transcript levels of genes related to sugar transport. The AL and transfer cell (TC) regions in the developing wheat caryopses at 9, 12, 15, and 19 days after anthesis were carefully separated by LCM (Supplementary Fig. 4), and RNA-seq was performed. Here, 77 expressed genes were screened from the transcriptome data of the TC vs. AL comparison. The number of expressed genes encoding sugar transporters was greatest at 9 DAA and decreased during the filling stage (Fig. 2a). The expression levels of most genes also decreased as the grains matured, indicating that the transport of sucrose was more active at the early filling stage (Fig. 2b). Additionally, the concentration of all sugars in the cavity sap peaked at 9 DAA, quickly decreased between 9 DAA and 12 DAA, and then continued to decrease until 19 DAA (Fig. 2d). This shifting pattern was consistent with the dynamic expression of carbohydrate transporters, which exhibited decreasing expression as anthesis progressed. Most genes were upregulated or specifically expressed in TC, especially at 15 DAA and 19 DAA, compared with those in AL (Fig. 2b). The sucrose transporter (SUT) and bidirectional sugar transporter SWEETs (SWEETs) are crucial for sucrose and glucose transport in plants, respectively, showing the highest expression among carbohydrate transporters. Notably, the expression of SUT and SWEET showed the same pattern in the TC and AL. The transcript levels of SUT and SWEET in the TC were much greater than those in the AL at all time points. The transcript levels of STP7, SPT, TPT, UXT2, IT2, UXT3, TMT3, and GPT2 were greater in the TC than in the AL at three or four-time points. The genes encoding URT2, PLT3, STP14, and PDPK2 exhibited high and specific expression (FPKM > 25) in TC. Together, these results support the essential role of transfer cells in sucrose transport.

Fig. 2: Spatial-temporal expression of carbohydrate transporter-encoding genes and sugar contents.
figure 2

a Venn diagram of unique shared carbohydrate transporter-encoding genes whose expression fold changes were ≥1.5 between the transferred cells and aleurone. b The colored boxes are heatmaps of gene expression related to carbohydrate transporters from RNA-seq, which were normalized by the log2 calculation method. For simplicity, homologous genes of the same gene were combined. Individual transcript levels, gene names, and IDs for each isoform can be found in Supplementary Data 2. c ALs and TCs of 7-DAA caryopses were dissected by LCM. The data are shown as the means ± SDs (n = 6). d Sugar contents in cavity sap. The data are shown as the means ± SDs (n = 3). SUT sucrose transporter, SWEET bidirectional sugar transporter SWEET, STP sugar transport protein, SPT probable sugar-phosphate/phosphate translocator, TPT triose-phosphate transporter domain-containing protein, UXT UDP-xylose transporter, TMT tonoplast monosaccharide transporter, GPT glucose-6-phosphate/phosphate translocator, IT probable inositol transporter, URGT UDP-rhamnose/UDP-galactose transporter, UTR UDP-galactose/UDP-glucose transporter, PLT probable polyol transporter, PDPK phosphoinositide-dependent protein kinase, G6PP glucose-6-phosphate/phosphate-translocator precursor, GONST GDP-mannose transporter, GFT GDP-fucose transporter, APE glucose-6-phosphate/phosphate translocator-like protein. Glu glucose, Suc sucrose, Fru fructose, Lac lactose, Mal maltose, Man mannose, All, allose.

Full size image

To quantify the sugar metabolites in the TC and AL, we dissected these tissues at 7 DAA before starch synthesis was initiated in the TC and AL (Supplementary Fig. 4). Total sugars were extracted and analyzed by LC‒MS. Seven sugars, namely, sucrose (Suc), glucose (Glu), mannose (Man), fructose (Fru), lactose (Lac), maltose (Mal), and allose (All), were profiled (Fig. 2c). Overall, the contents of all sugars were greater in the TC treatment than in the AL treatment, with the exception of Fru, the contents of which were similar in the two tissues. Few allose and mannose were detected in AL, while their contents were relatively high in TC. The sucrose contents in both tissues were much greater than those of the other sugars, confirming that sucrose was the major sugar transported from the mother plant to the caryopses. The sucrose content in the TC treatment group was 1.98 times greater than that in the AL group, which verified the importance of TC in sucrose transport.

Gene set enrichment analysis (GSEA) reveals differentially expressed transcripts between the aleurone (AL) and outer endosperm (OE)

After elucidating the sucrose transport pathway, we focused on the transcriptomic variations of cells in different regions to investigate how differences in starch content occur. The AL, OE, and inner endosperm (IE) of the developing wheat caryopses at 9, 12, 15, and 19 days after anthesis were carefully separated by LCM (Supplementary Fig. 4), and 36 complementary DNA (cDNA) libraries were constructed. After sequencing, the total number of mapped reads ranged from 44,919,184 to 66,087,412, with the percentage of unique mapped reads ranging from 65.06% to 76.54% (Supplementary data 1). There were 4895–14,978 genes detected, and the number of DEGs between samples is shown in Supplementary Fig. 5.

To identify the key pathways altered in AL, we performed GSEA on the transcriptomic data of AL vs. OE patients. The results showed that the starch and sucrose metabolism, biofilm formation, and biosynthesis pathways were downregulated in AL (p < 0.01). In comparison, the metabolism of amino sugars and nucleotide sugars, peroxisomes, and glutathione in AL plants was more regulated than that in OE plants (p < 0.01) at 19 DAA. To critically identify the specific metabolic pathways involved in starch synthesis, we focused on the starch and sucrose metabolism dataset. The enrichment plot reflects the degree to which a gene set is overrepresented at the top or bottom of a ranked list of genes. The score at the peak of the plot (the score farthest from 0.0) is the enrichment score for the gene set. The vertical lines indicate the positions of individual members of the gene set in the ranked list. These data suggested that most genes involved in starch and sucrose metabolism were downregulated in AL plants compared with those in OE plants (Fig. 3 and Supplementary Fig. 6), which was consistent with the much lower starch content in AL plants (Table 1). No significant gene enrichment in starch or sucrose metabolism was detected in IE vs. OE (Supplementary Fig. 6), which was also in accordance with the small difference in starch content between IE and OE. We further investigated this pathway by exploring the spatial dynamics of sucrose metabolism and starch synthesis capacity.

Fig. 3: Gene Set Enrichment Analysis (GSEA) of transcriptomic data from aleurone vs. outer endosperm.
figure 3

a Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of aleurone compared with the outer endosperm of 19-DAA caryopses. b GSEA showing a high enrichment score (ES) of the KEGG term starch and sucrose metabolism in samples from the aleurone vs. outer endosperm.

Full size image

Expression of sucrose metabolism- and starch synthesis-related genes indicating the importance of the conversion capacity of sucrose to starch in the spatial starch gradient in the wheat endosperm

After confirming the relationship between substrate supply and starch gradient, we investigated genes related to sucrose metabolism and starch synthesis. In total, 181 genes involved in sucrose metabolism and 27 genes involved in starch synthesis were screened by KEGG pathway enrichment.

A schematic diagram of sucrose metabolism is shown in Fig. 4a. Genes involved in sucrose metabolism were analyzed and are presented on a log2 scale as relative transcript levels in different tissues at 9, 12, 15, and 19 DAA (Fig. 4a). For simplicity, members of gene families were combined at each metabolic step, and individual transcript levels, gene names and IDs for each isoform can be found in Supplementary data 2. Overall, most genes exhibited differential expression among AL, OE, and IE tissues at all time points, especially in AL, compared with the other two tissues. Furthermore, many genes showed substantially differential expression levels in the same tissue among different time points. The expression of most genes was greatest at 9 DAA and then decreased during grain filling.

Fig. 4: Spatio-temporal variations in the expression levels of genes involved in sucrose and starch metabolism.
figure 4

a The colored boxes are the heatmaps of gene expression data for sucrose metabolism obtained from RNA-seq, which were normalized by the log2 calculation method (Supplementary data 2). RNA was extracted from the aleurone (AL), outer endosperm (OE), and inner endosperm (IE) of 9-, 12-, 15-, and 19-day development caryopses, with three replicates for each. For simplicity, members of gene families were used. b The difference in the number of RPKMs of genes encoding starch biosynthesis between the AL or IE group and the OE control group was calculated (means ± SDs, n = 3). Three bars at the same time point represent homologous genes of the same gene. RPKM reads per kilobase per million mapped reads, S6PH sucrose-6F-phosphate phosphohydrolase, SPS sucrose-phosphate synthase, F6T ATP:D-fructose 6-phosphotransferase, G6PI D-glucose-6-phosphate aldose-ketose-isomerase, G6T ATP:D-glucose 6-phosphotransferase, G1,6P alpha-D-glucose 1,6-phosphomutase, β-GH beta-D-glucoside glucohydrolase, α-GH alpha-D-glucoside glucohydrolase, α-G1PT ATP:alpha-D-glucose-1-phosphate adenylyltransferase, α-G1PU UTP:alpha-D-glucose-1-phosphate uridylyltransferase, G6PT UDP-glucose:D-glucose-6-phosphate 1-alpha-D-glucosyltransferase, Susy sucrose synthase, α-GT 4-alpha-glucanotransferase, 4α-GH 4-alpha-D-glucan maltohydrolase, 3β-GH 3-beta-D-glucan glucanohydrolase, 4β-GH 4-beta-D-glucan 4-glucanohydrol. The individual transcript levels, gene names, and IDs for each isoform can be found in Supplementary Data 2.

Full size image

The first step in sucrose degradation is catalyzed by sucrose synthase (Susy) and alpha-glucosidase (α-GH) to form UDPG, fructose, and glucose. There was a close relationship between the different expression levels of genes related to sucrose degradation and starch spatial differences. For example, the transcript levels of both Susy and α-GH were lower in AL than in OE and IE. In the endosperm, the transcript levels of sucrose synthase were greater in IE than in OE at 9 DAA and 19 DAA, indicating that there might be a relatively greater tendency toward carbohydrate accumulation in IE.

UDPG could contribute directly to glucose formation or indirectly through cellobiose or maltose biosynthesis. The expression levels of most genes participating in these pathways were much lower in the AL than in the endosperm, especially at 9 DAA. UDPG could also support the biosynthesis of α-D-glucose-1P by alpha-glucose-1-phosphate adenylyltransferase (α-G1PT). The transcript levels of α-G1PT were similar in the three tissues at 9 DAA, whereas the levels in the ALs were lower than those in the OEs at 12, 15, and 19 DAA. Compared with that in the OE, α-G1PT exhibited greater expression in the IE at 12 DAA, while it had lower expression in the IE at 19 DAA. Thus, these results supported a transcriptional downregulation in UDPG metabolism to explain the low starch accumulation in the AL.

Another pathway for the production of α-D-glucose-1P involves fructose and glucose metabolism, which is catalyzed by fructose 6-phosphotransferase (F6T), glucose 6-phosphotransferase (G6T), glucose-6-phosphate 1-alpha-glucosyltransferase (G6PT), sucrose-phosphate synthase (SPS) and glucose 1,6-phosphomutase (G1,6 P). Interestingly, the spatial distribution patterns of genes encoding these enzymes were the opposite of those described above. The transcript levels of F6T were greater in the AL than in the endosperm at all time points. The expression of genes encoding SPS, G6PI, and G6T was greater at 9 DAA but lower at 15 DAA in the AL than in the endosperm. In addition, the genes encoding SPS and G1,6P were upregulated in the AL. Together, the expression of genes participating in fructose and glucose metabolism was generally greater in the AL, and the expression levels of these genes were lower in the caryopses than in the other genes, so it is unlikely that these steps were important for regulating the spatial distribution of starch.

The final step before starch biosynthesis in cereals is the formation of ADP-glucose, a glucose donor, from glucose 1-phosphate (G1P) and ATP, which is catalyzed by G1P adenylyltransferase (α-G1PT). The genes encoding α-G1PT were downregulated significantly in response to aleurone at all time points. These genes exhibited significantly greater expression in IE than in OE at 9 and 19 DAA, which was consistent with the observation that IE had greater amounts of starch.

Since there is a close relationship between starch synthase and starch content, the transcriptional differences in starch synthesis genes are worth explicitly noting. The transcript levels of these genes in the OE were chosen as the control, and the relative variations in AL and IE in relation to those in the OE were calculated (Fig. 4b). Three homologous genes encoding starch synthase and starch branching enzymes at different days after anthesis were evaluated. All these genes were significantly downregulated in the AL vs. OE comparison, with the exception of DBE, the expression of which decreased significantly only at 9 and 15 DAA. The differences between IE and OE were much narrower than those between AL and OE. The transcript levels of GBSSI, SSI, SSII, and SBEIIa were significantly greater in the IE group than in the OE group at 9 and 12 DAA for some of the homologous genes. An obvious upregulation was observed in the homologous gene on 1D encoding SBEI in IE vs. OE. The expression of GBSSI was lower in IE plants than in OE plants at 15 and 19 DAA. There was a relatively greater tendency toward DBE transcripts in IE, and the differences were mostly insignificant.

Validation of key differentially expressed genes

To validate the RNA-seq data, the transcription expression of nine key DEGs involved in sucrose transport (Fig. 5a), sucrose metabolism (Fig. 5b–e), and starch synthesis (Fig. 5f–i) was analyzed by qRT‒PCR. LCM revealed the AL, OE, IE and TC of caryopses at 9 DAA (Fig. 5) and 15 DAA (Supplementary Fig. 7). The gene encoding the sucrose transporter was expressed at the highest level in the TC, followed by the IE, and was expressed at the lowest level in the OE and AL, confirming that the sucrose transport pathway involved the transfer of cells from the inner endosperm to the outer endosperm. The transcript levels of genes involved in sucrose metabolism and starch synthesis were greater in the endosperm than in the AL, consistent with the spatial distribution of starch in mature grains. Overall, the expression trends of the nine genes according to qRT‒PCR were highly consistent with those according to the RNA-seq results, with correlation coefficients between 0.95 and 0.99 (Fig. 5).

Fig. 5: Comparison of mRNA expression patterns of nine representative genes by qPCR (left) and RNA-seq (right).
figure 5

a Sucrose transporter. b G1P adenylytransferase. c Sucrose synthase. d Endoglucanase. e Beta-amylase. f GBSSI. g SSI. h SSII. i SBE. mRNAs were extracted from cells of aleurone, outer endosperm, inner endosperm, and transfer cells of caryopses at 9 DAA. The gene IDs and primers used are listed in Supplementary Data 3.

Full size image

To further confirm the spatial expression patterns of some representative genes, we used RNA fluorescence in situ hybridization (FISH) for direct visualization of the spatial expression patterns of genes encoding SUT, SWEET, and β-amylase. Both strong red (SUT transcript) and green (SWEET transcript) RNA FISH signals were observed in both the nucellar and transfer cells (Supplementary Fig. 8B, C). In contrast, no apparent β-amylase signal was detected in the nucellar or transfer cells (Supplementary Fig. 8D). In the inner endosperm, only a strong signal for β-amylase mRNA was observed (Supplementary Fig. 8E–G). Taken together, these findings indicate that both carbohydrate transporters were more highly expressed in the transferred cells, whereas β-amylase was more highly expressed in the inner endosperm, which is consistent with the transcriptomic and qRT‒PCR results.

Discussion

Several reports have described the variation in starch composition within the starchy endosperm. Here, we confirmed that the starch content in aleurone was only half of that in endosperm, consistent with previous reports that only small proportions of starch are present in bran and aleurone layers4. In terms of starch components, the amylose contents were similar in the OE and IE, while the amylopectin content was greater in the IE than in the OE, which resulted in a decrease in the amylose/amylopectin ratio in the IE. However, Tosi et al.2 reported that the amylose/amylopectin ratio increased inwardly and reached a maximum at P6 or P72. This inconsistency was probably a consequence of the use of different pearling methods, in which the grains were abraded into seven fractions (7%, 6%, 7%, 10%, 10%, 10%, and 50% of the grain weight), and our IE corresponded to part of the core fraction of that study. The differences in the starch composition of the pearling fractions are expected to influence the functional properties of the flours. Here, we observed gradients in pasting properties, which was consistent with a previous report4.

Although the starch gradients in wheat grains are clear, the mechanisms determining the establishment of such gradients during wheat grain development are quite limited. The substrate supply or the genes encoding starch synthesis, or both factors attributed to the formation of starch gradients, remain to be explored. In the present study, we constructed a framework for understanding sucrose transport and starch synthesis in different parts of caryopses at different filling stages via transcriptomic analysis. Overall, a range of 4895–14,978 genes were identified in various samples. These numbers were lower than those in a previous study, which revealed a progressive decrease in the number of expressed genes from 23,523 to 13,722 between 5 DAA and 20 DAA. The limited collection of genes detected might be attributed to the separated caryopsis sections. Starch synthesis commences with the transportation of sugar from mother plants into the endosperm, facilitated by carbohydrate transporters23. In the present study, 74 genes encoding carbohydrate transporters were expressed at 9 DAA in TC, and abundant sugars were detected in TC at 7 DAA (Fig. 2a), indicating that this tissue provided substrate for starch synthesis during the early filling stage. Thiel et al.24 conducted a microdissection-based GC‒MS study of nucellar projection and transfer cells in barley. The authors reported that the contents of most sugars were the highest at 5 or 7 DAA in both tissues24. After being transported to the endosperm, sucrose is converted into other sugars and finally to ADPG5, which is crucial for providing substrates for starch synthesis. A schematic diagram of sucrose metabolism clearly showed that the genes related to sucrose conversion were most highly expressed at 9 DAA, and their expression decreased gradually during the filling stage (Fig. 4a). This observation indicated that sucrose conversion was activated immediately after the endosperm received sugar from the TC. Afterward, starch synthesis was carried out by the orchestrated action of all starch synthases. Here, we found that the transcripts for starch synthase were most abundant at 12 and 15 DAA (Fig. 4b), which was in agreement with previous reports that a rapid increase in starch content started at 12 DAA7. To improve our knowledge of the mechanism of starch gradient formation, sucrose transport, sugar metabolism, and starch synthesis were comprehensively discussed from a spatial perspective.

The nutrients transported from the vascular bundle were first unloaded to the apoplastic space via chalaza and nucellar projections and then transported into the endosperm via two pathways (Supplementary Fig. 1). This study confirmed that the substrate for starch synthesis is transported by the first pathway mentioned above, with the following three pieces of evidence. First, we obtained the direct and nondestructive transport dynamics of FDG via PET-CT (Fig. 1). The FDG gradients presented in the sections of caryopses demonstrated the transport pathway from the transferred cells to the endosperm. PET has been used extensively for human and animal research for more than 30 years. It is commonly used to trace the uptake and metabolism of D-glucose in metabolically active tissues, such as brain tissue or cancer cells. However, it is rarely used in plant imaging studies to trace sugar dynamics25. There is a concern that the transport pathway of [18F]FDG will be different from that of sucrose. However, the in situ RNA evidence in our study revealed identical expression patterns between the sucrose transporter (SUT) and glucose transporter (SWEET) genes (Supplementary Fig. 8). Both exhibit high expression in transfer cells and nucellar projections, suggesting a parallel transport route for both sugars. It has been reported that [18F]FDG could be used as a model of photoassimilates to study sugar uptake and distribution in giant reed plants16. In addition, some genes encoding uniporters that mediate the efflux of glucose, such as AtSWEET1 and ERD6-like 10, both interact with AtSUCs (sucrose transporters) in Arabidopsis23, indicating that there might be crosstalk between sucrose and glucose transport pathways; this also explained why glucose-fed spikes developed well in our study (Supplementary Fig. 2). Overall, the transport pathway of the starch substrate in developing caryopses can be confirmed, at least in part, by direct observation of [18F]FDG via PET-CT.

The second example of evidence is the distribution of metabolites in transfer cells and aleurone at the early filling stage before starch synthesis is initiated in these two tissues. Most of the sugars, especially sucrose, were more abundant in the TC treatment than in the AL treatment (Fig. 2c). This result was consistent with a previous report showing that oligofructan accumulation in the endosperm cavity was highest in the cell layers surrounding this cavity26. More recently, Verhertbruggen et al.27 strongly detected mannan in the walls of the first cell layers bordering the endosperm cavity, and a gradient of signal intensity was observed from the ventral to the dorsal region of the endosperm by immunolabeling27, which was also consistent with our results. Additionally, the glucose content showed the same distribution as that of sucrose, which also supported the PET-CT observations. The metabolites in the OE and IE were also measured, but the data were irregular since starch was synthesized in the endosperm at 7 DAA (Supplementary Fig. 4).

Third, this sucrose transport pathway was also supported by the finding that genes encoding sugar transporters are differentially expressed and are more strongly expressed in the TC than in the AL (Fig. 2b). The expression of genes encoding either sucrose transporters (SUTs) or glucose transporters (SWEETs) was 2.66- to 25.44-fold greater in the TC treatment than in the AL treatment, as indicated by the transcriptomic data. Thiel et al.24 reported that TC is the main site of solute uptake into the endosperm, as suggested by the transcriptional activity of assimilate and micronutrient transporters24. According to the transcriptomic data, the expression levels of SUT were the highest in TC, followed by those in IE, and were the lowest in both OE and AL, as indicated by qPCR. In conclusion, the sucrose transport pathway in caryopses was TC—IE—OE—AL.

Based on the finding that IE can preferentially acquire substrates for starch synthesis, it is reasonable to suspect that more active sucrose metabolism occurs in IE than in the other two tissues. In this study, while most differentially expressed genes between IE and OE plants were not enriched in sucrose metabolism, as shown by GSEA (Supplementary Fig. 6), some core genes, such as sucrose synthase, G1P acetyltransferase and beta-amylase, still exhibited significantly greater expression in IE plants than in OE plants, as confirmed by both transcriptomic data (Fig. 4a) and qPCR data (Fig. 5). This result is consistent with a previous report that inulin-type fructans appear to be biosynthesized and accumulate more in cells surrounding the endosperm cavity26. Most of the sucrose metabolism pathways were significantly downregulated in AL, which is consistent with transcripts that promote the rapid accumulation of carbohydrates more frequently detected in endosperm than in aleurone28. Additionally, the genes involved in fructose and glucose metabolism were more highly expressed in AL (Fig. 4a), which might compensate for the increase in the production of α-D-glucose-1P.

Another marked difference that could contribute to the starch gradients in caryopses was the transcriptional capacity for starch synthesis. Most of the genes encoding starch synthase showed lower expression in AL than in IE and OE (Fig. 4b), which may have resulted in the low starch content in AL. It should be noted that as the furthest tissue from the cavity and transfer cells, the substrate supply in AL could be lower than that in the endosperm. This finding raised the question of whether the differences in expression among the AL, OE, and IE treatments were due to differences in substrate supply capacity. It has been reported that overexpressing SSIV results in increased reserve starch levels and yield in potato tubers; however, these results are accompanied by increased ADPG29. Except for this paper, the ectopic expression of SS genes has rarely been used to increase the accumulation of starch because the SSs in plants interact with each other in a protein complex to regulate the final architecture of the starch granule30. From this perspective, it might be easier to increase the starch content in AL by enhancing the sugar supply. However, the substrate supply difference is certainly not the reason for the higher amylopectin content in IE than in OE. Since amylose and amylopectin synthesis share the same substrate, the higher content of amylopectin in IE could only be due to higher SSII and SBEIa expression (Fig. 4b). Therefore, starch synthase could be the only target for manipulating the ratio of amylose/amylopectin inside the endosperm.

Some studies have focused on the modulation of component gradients in the endosperm by nutrients or the environment. He et al.31 reported that increasing nitrogen availability increased the proportions of ω-gliadins in all pearling fractions. The proportion of HMW subunits increased in all fractions except the core31. Similarly, Zhong et al.32 studied the effects of nitrogen fertilization on protein distribution, showing that it could be modulated by the timing of application of topdressing32. Additionally, the gradients could also be modified by the environment, such as temperature. Savill et al.33 reported that protein gradients were enhanced by high temperatures postanthesis33. However, few breeding strategies for modifying component gradients in wheat endosperm have been developed; this is understandable because the spatial distribution of the components was not ascribed to one or two genes. One successful case has been conducted in barley34. The authors overexpressed the vacuolar Zn transporter HvMTP1 under the control of the endosperm-specific D-hordein promoter and observed a redistribution of grain Zn from the aleurone to the endosperm in the transformed plants. In this study, we confirmed that starch gradients are closely related to sucrose supply and that the sucrose transporter is responsible for the sucrose supply in grain. Thus, changing the starch gradient via a strategy similar to that used by Menguer et al.34 is promising. Additionally, although studies on modifying starch gradients in wheat endosperm are quite limited, it has been reported that overexpressing the cytosolic AGPase large subunit gene35 or sucrose synthase gene36 can increase the starch content and grain weight of common wheat. We also observed that the expression of these genes was closely related to the final starch distribution in the endosperm. Thus, these genes could be potential targets for modifying starch gradients through the coordination of suitable promoters.

In summary, a model has been defined for starch gradient formation in wheat caryopses based on the sucrose transport pathway acquired by PET-CT and gene expression profiles of laser microdissected tissues (Fig. 6). Sucrose, the substrate for starch synthesis, is transported radially from the cavity to the endosperm via transfer cells. In this case, the inner endosperm could acquire sucrose preferentially over the aleurone. The genes involved in sucrose metabolism and starch biosynthesis were therefore downregulated significantly compared with those in the outer endosperm, resulting in a lower starch content and smaller starch granules in aleurone. We would like to emphasize that the combination of genes in the same metabolic step may provide some information, so the transcript levels presented in Supplementary Data 2 should be an important database for future studies on regulating starch gradients in wheat grain. Furthermore, consideration of the endogenous heterogeneity of starch metabolism will be economical for optimizing the starch outcomes of wheat grains.

Fig. 6: Proposed models of factors contributing to starch gradients in wheat caryopses.
figure 6

The sucrose concentration is presented as the number of orange circles modified from the PET-CT results (Fig. 2). The black arrows indicate the transport pathway of sucrose in the caryopsis from the cavity to the aleurone via the transferred cells. The number of green cubes is used to indicate the expression levels of sucrose transporters according to the qRT‒PCR results (Fig. 5). Sucrose metabolism and starch synthesis ability in different parts of the caryopsis are indicated by the thickness of the red arrows. The contents of amylose and amylopectin are shown as different shades of blue. The transverse section is from the developing grain of wheat at 15 days. AL aleurone, OE outer endosperm, IE inner endosperm, TC transfer cells, SG starch granule, s sucrose. Scale bar = 30 μm.

Full size image

Methods

Plant materials and growth conditions

Plants were grown in pots placed in an artificial climate chamber under a day/night temperature regime of 24/16 °C and a 12-h photoperiod at Nanjing Agricultural University, Nanjing (32°08′N and 118°51′E), China, in 2019. Pots with a depth of 18 cm and diameter of 20 cm were filled with 4.5 kg of clay soil. A local broadly grown winter wheat cultivar (Triticum aestivum L. cv. Yangmai 16) was planted with 4 plants per pot. Before sowing, 1 g of urea and 0.8 g of KH2PO3 were applied to each pot. Another 1 g of urea per pot was top-dressed at the jointing stage.

Grain pearling

Mature-grain pearling was generated following our previous report with minor modifications32. Briefly, 20 g grains of uniform size were picked manually. Three pearling fractions were sequentially prepared using the Foodstuff Machine Pearlest (Kett Electric Laboratory, Japan). The pearling fractions were collected as flour enriched in the husk and aleurone layer (the P1, 20%), outer endosperm (the P2, 35%), and inner endosperm (P4, 35%). The remaining 10% of the grains were discarded to avoid the effect of grain cavities. Each pearling fraction was sieved through an 80-mesh sieve.

Starch, amylose, and amylopectin quantification

The amylose and amylopectin contents of each pearling fraction were determined with a coupled spectrophotometer assay37. Hundred milligrams of flour were prepared for analysis and 2 ml of KOH solution (0.5 M), which was diluted to a volume of 50 ml with distilled water after keeping for 20 min at 85 °C, reacted with the I2-KI reagent. Amylose was measured using an absorbance of 610 nm (OD620) and a reference wavelength of 490 nm. Amylopectin content was measured using an absorbance of 550 nm (OD550) and a reference wavelength of 682 nm. All the measurements were repeated three times. The amylose and amylopectin standards were purchased from Sigma–Aldrich (St. Louis, USA).

The starch content was determined by polarimetry method38. One gram of flour was dissolved by boiling in 50 mL dilute hydrochloric acid (1.125% HCl), dissolved proteinaceous substances were precipitated with 5 mL Carrez I solution and with 5 mL Carrez II solution and then filtered. The optical rotation in angle degrees (α) of the filtrate was measured and total starch (TS) content is calculated as:

$${{\rm{TS}}} \% =\frac{2000 \, {{\cdot }} \, {{\rm{\alpha }}}}{{[{{\rm{\alpha }}}]}_{D}^{20}}$$

The [α]D20(°) for wheat flour is 182.7.

Measurement of pasting characteristics

The pasting properties of each pearling fraction were analyzed using a Super 3 Rapid Visco Analyzer (Newport, Australia) following the AACC method 76–21 (AACC, 2000). Three and a half grams of flour (the weight was adjusted based on 14% moisture) was suspended in 25 ml of deionized water in a measuring cup. The viscosity was then measured using a Rapid Visco Analyzer. The embedded standard protocols for measuring the viscosity of wheat flour were selected.

PET-CT

PET-CT imaging studies were performed on in vitro spikes using an Inveon small-animal PET-CT scanner (Siemens). First, stems were cut to approximately 15 cm (panicle not included) and over the first node to exclude all leaves and then transferred to 15 ml plastic tubes with media (0.5 g l−1 2[N-morpholino] ethane sulfonic acid, 0.8 g·l−1 glutamine, and 1.47 g l−1 Murashige-Skoog medium). [18F]FDG (300 μCi) was injected into the media, and the tubes with spikes were placed into the scanning chamber, as shown in Fig. 1. Then, imaging acquisition started with a low-dose CT scan, followed by a 10 min per bed position PET scan immediately. The CT scan was used for attenuation correction and localization of the spike. PET image acquisition was performed every half minute in 3-dimensional mode for 90 min. As a control, one sample was scanned via micro PET-CT without [18F]FDG injection. The images were reconstructed, and the regions of interest (ROIs) were drawn over the whole spike images using Inevon Research Workplace 4.1 software.

Cryosection and laser capture microdissection (LCM)

Cryosectioning and LCM were conducted using a modified version of the protocol described by Tauris et al.39. Ears flowering on the same day were tagged, and the caryopses in the middle of the tagged ears were collected at 9, 12, 15, and 19 days after anthesis (DAA). The fresh caryopses were embedded in Neg-50 Frozen Section Medium (Richard-Allan Scientific) immediately after collection, sectioned at 30 µm in a cryostat at −30 °C and mounted on FrameSlides (Leica, Germany). The slides were dipped in 100% ethanol to dehydrate the sections completely before the laser capture microdissection procedure. The cells of the aleurone, outer endosperm, inner endosperm, and transferred cells were excised with a Leica LMD7000 system. Then, total RNA was extracted with the Arcturus PicoPure RNA Isolation Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The integrity of the RNA was analyzed on a 2100 Bioanalyzer (Agilent Technologies) before being subjected to qRT‒PCR or transcriptome analysis (Supplementary Fig. 9).

Construction of RNA-seq libraries

Total RNA was used to construct RNA-seq libraries according to the manufacturer’s protocol (N712, Vazyme, China). Considering the low amount of RNA obtained from LCM samples, 16 cycles were performed to obtain the final cDNA library with an average insert size of 350 bp (±50 bp). Then, we performed paired-end sequencing on an Illumina HiSeq X-Ten (LC Bio, China) following the vendor’s recommended protocol.

For bioinformatics analysis, Cutadapt was first used to remove the reads containing adapter contaminants, low-quality bases, and undetermined bases40. Then, sequence quality was verified using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). The reads were mapped to the wheat genome IWGSC in Ensemble 41 by HISAT241. The mapped reads of each sample were assembled using StringTie42. Then, the transcriptomes from all the samples were merged to reconstruct a comprehensive transcriptome using Perl scripts. Subsequently, taking sequencing depth, gene length, and read counts into consideration, the expected number of fragments per kilobase of transcript sequence per million base pairs sequenced (FPKM) was calculated to quantify each gene expression level43. The differentially expressed mRNAs with log2 (fold change) >1 or log2 (fold change) <−1 and with statistical significance (p < 0.05) were selected by the R package edgeR44. Gene Ontology (GO) and Kyoto Encyclopedia of Genes (KEGG) pathway analyses were performed using DAVID 6.7 or Cytoscape with the ClueGO plug-in. For the identification of enriched pathways, we used the gene set enrichment analysis (GSEA) tool (v3.0) from the Broad Institute at the Massachusetts Institute of Technology. GSEA was performed by comparing the normalized gene expression data obtained from aleurone and the outer endosperm at the same stage.

Analysis of spatial gene expression by qRT‒PCR

cDNA was synthesized from LCM RNA samples using the PrimeScript RT reagent Kit (Takara, Japan). Power SYBR Green reagent (Vazyme, China) was used with a LightCycler 96 for qRT‒PCR (Roche, Switzerland). Details of the primers used are listed in Supplementary Data 3. All primers were gene-specific, as determined by electrophoresis and unique melting peak verification. Three biological replicates were analyzed in triplicate for each PCR for each gene.

Metabolite extraction and LC‒MS

The cells of aleurone and the transferred cells of caryopses at 7 DAA were excised and collected in tubes (0.5 mL) using the abovementioned method (see MM 4.6). All the samples were accurately weighed to 0.01 mg. The cavity sap of the endosperm at 9, 12, 15, and 19 DAA was collected by slicing off the hilum end of the grain, sucking the fluid with a capillary, and pooling the fluid of thirty grains for each replicate44. Dissected cells or cavity sap were dissolved in 700 μl 80% ethanol, vortexed thoroughly, and incubated at 50 °C for 2 h. The mixture was added 700 μl water with centrifuged at 10,000 × g for 3 min, and the supernatant was collected into a new tube for HPLC analysis45.

RNA Fluorescent in situ Hybridization (FISH)

For synthesis of SUT, SWEET, and β-amylase probes, gene-specific primer sets as shown in supplementary data 3, were used. Developing caryopsis of 9 DAA was collected, fixed, dehydrated, embedded, tissue section to 10 μm, and performed in situ hybridization. Tissue samples were dewaxed in xylene for 10 min each 2 times, and then were re-hydrated in 100%, 85%, and 75% ethanol for 3 min each, followed by in PBS for 3 min. Afterward, the samples were protease digested and prehybridized as described by Jandura et al.46. After that, three probes were mixed with a SUT probe (C1-red), SWEET probe (C2-green) and β-amylase probe (C3-cyan) diluted to a hybridization buffer in a ratio of 1:1:1:50, denaturized at 85 °C for 3 min and equilibrated at 37 °C for 5 min. Then, added the probe dropwise to the tablet, completely cover the tissue, and place in the hybridizer to hybridize overnight at 37 °C. After being washed for 3 times, slides were counterstained with 4′, 6-diamidino-2-phenylindole (DAPI). Florescence signals were observed using a fluorescence microscope (Nikon 80i, Japan).

Statistics and reproducibility

All the data were subjected to analysis of variance (ANOVA) using SPSS (Statistical Product and Service Solutions) Version 17.0. ANOVA mean comparisons were performed in terms of the least significant difference (LSD) at the significance level of P < 0.05. Heatmap was visualized by R package ‘pheatmap’ in R 3.5.1. Bar charts were visualized by GraphPad Prism.

For starch and starch composition analysis, three independent biological replicates were carried out. For RNA-seq analysis, each sample included three independent biological replicates. In qPCR, three biological replicates and three technical replicates were carried out. For sugar quantification of cavity sap, we used three biological replicates while for sugar quantification of dissected aleurone and transfer cells, five biological replicates were carried out.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.