Functional definition of leaf sink-to-source transition by ¹⁴C tracing reveals transcriptional reprogramming in soybean

Abstract

Immature leaves initially act as sinks that import photosynthates from mature leaves. As leaves develop, they gradually acquire source capacity—a developmental shift known as the sink-to-source transition. Although this transition is critical for increasing source number and yield potential, its molecular basis remains largely unclear. To investigate the underlying mechanisms in soybean, we combined carbon-14 (¹⁴C) tracer analysis with RNA-seq. At an early vegetative stage, the second trifoliate leaf was classified into sink, transition, or source states based on its ¹⁴C accumulation pattern. Autoradiographic analyses further indicated that regulation of photosynthate transport occurs at the nodal region, prompting parallel RNA-seq analysis of both leaf and node tissues. Transcriptome analysis revealed early induction of cytokinin negative regulators, A-type ARABIDOPSIS RESPONSE REGULATORs (ARRs), in leaf tissue, suggesting attenuation of cytokinin signaling at the onset of the sink-to-source transition. In contrast, genes involved in sugar transport, such as sugars will eventually be exported transporters and sucrose transporter family members, showed continuous upregulation across developmental stages in both leaf and node tissues. Together, these results provide a high-resolution view of transcriptional reprogramming accompanying the leaf sink-to-source transition in soybean.

Introduction

Plants produce sugars through photosynthesis and use them to sustain growth, development, and reproduction. How these sugars, or photosynthates, are distributed among plant organs strongly affects plant performance and yield. As they develop, leaves shift from net sugar-importing “sinks” to net sugar-exporting “sources.” This developmental change, known as the sink-to-source transition, marks the onset of net carbon export and plays an essential role in increasing source number and yield potential.

Classical physiological studies using radioactive ¹⁴C tracers revealed that sink and source regions can coexist within a single leaf. Leaf maturation proceeds basipetally from the tip toward the base, with phloem loading and unloading occurring mainly in minor and major veins, respectively (Fellows and Geiger 1974, Turgeon 1987, 1989). In higher plants, the phloem serves as the principal pathway for long-distance transport of photosynthates, with sucrose being the predominant form of these assimilates. Within leaves, sucrose moves between cells either symplastically through plasmodesmata (PD) or via apoplastic pathways mediated by membrane transporters. Developmental remodeling of PD from simple to branched forms coincides with this transition and is linked to changes in phloem assimilate transport (Roberts et al. 2001, Tee and Faulkner 2024).

Over the past decades, molecular research has elucidated the key transporters and enzymes and regulatory signals underlying this transition. Sink tissues, which require imported sugars to sustain their growth and metabolism, receive photoassimilates through phloem unloading. This unloading occurs either via the symplastic pathway through PD or via the apoplastic pathway, where released sucrose is directly taken up by sucrose transporters (SUTs) or first hydrolyzed by cell wall invertase (CWINV) before hexose uptake (Stein and Granot 2019). The balance between symplastic and apoplastic pathways shifts dynamically during development (Braun 2022). In source leaves, excess sucrose is loaded into the phloem via apoplastic and/or symplastic pathways, depending on species. Apoplastic pathway involves Sugars Will Eventually be Exported Transporters (SWEETs) that mediate sucrose efflux to the apoplast, followed by reuptake into the phloem through SUTs.

Sink and source tissues show contrasting transcriptional profiles. Sink leaves typically exhibit high expression of sucrose synthase (SUS) and invertase (INV) genes (Stein and Granot 2019, Liu et al. 2025), supporting sucrose cleavage and downstream metabolism, whereas source leaves upregulate genes associated with photosynthesis and sugar biosynthesis (Li et al. 2022). Hormone-responsive genes also undergo marked developmental regulation, linking leaf maturation not only to carbon allocation but also to growth and stress response (Ribeiro et al. 2021). Together, these findings indicate that the sink-to-source transition is governed by an integrated regulatory network that coordinates sugar transport, metabolism, and signaling.

Despite this progress, most transcriptomic studies have focused on comparing only young and mature leaves. The intermediate transition stage has rarely been clearly defined, and its fine-scale transcriptional dynamics are often overlooked. This transitional window is critical because it marks the stage when leaves acquire source function and begin to export carbon on a net basis. Although several studies have reported gene expression changes during the leaf maturation period (e.g. SUS downregulation in maize: Cheng et al. 1996; Rubisco upregulation in amaranth: Wang et al. 1993; and primary cell wall-related gene expression in rice: Joo et al. 2021), these observations remain fragmented and unintegrated. A comprehensive genome-wide analysis of the transcriptional reprogramming that defines this transition has yet to be conducted, with sugar transport activity serving as one of its most informative functional markers.

Building on these observations, recent large-scale “cell atlas” studies in Arabidopsis and soybean have begun to reveal transcriptional networks underlying nutrient transport and tissue-specific regulation (Guo et al. 2025, Zhang et al. 2025). In Arabidopsis, senescence involves coordinated transcriptional programs that mobilize carbon and nitrogen (Guo et al. 2025), whereas in soybean, multiple sugar transporters share binding motifs for DOF transcription factors, suggesting conserved regulatory modules (Zhang et al. 2025). These findings highlight the central role of transcriptional regulation in maintaining source–sink balance; however, how these networks operate specifically during the sink-to-source transition remains to be elucidated.

To our knowledge, this study represents the first systematic effort to resolve genome-wide transcriptional reprogramming precisely at the functional transition point. Here, we addressed this gap by characterizing the transcriptional dynamics of the sink-to-source transition within a high-resolution spatiotemporal framework. By combining ¹⁴C tracer autoradiography with RNA-seq in soybean (Glycine max), we quantitatively defined the transition zone based on photosynthate transport activity and captured gene expression changes across narrow developmental windows. Soybean was chosen not only for its global agricultural importance as a major protein and oil crop (Hartman et al. 2011, Liu et al. 2020), but also because its large leaf size and distinct vein architecture facilitate fine-scale sampling and autoradiographic imaging. Autoradiography further revealed that transport regulation also occurs in stem tissues, prompting us to extend our analyses beyond leaves. By integrating ¹⁴C autoradiography with transcriptome profiling, we provide a comprehensive view of the molecular reprogramming that accompanies the sink-to-source transition. The resulting dataset establishes a foundation for identifying regulatory factors that control carbon allocation and may guide future strategies to enhance carbon-use efficiency and yield potential in soybean and other crops.

Results

Definition and classification of second trifoliate leaf sink–source states

In this study, we defined “sink,” “transition,” and “source” leaves based on their ¹⁴C allocation patterns and export activity. Using this allocation-based classification enabled us to determine the precise developmental status of the second trifoliate leaf (L2) prior to RNA-seq.

To determine the sink, transition, and source states of L2, we characterized its growth progression and the distribution patterns of ¹⁴C-photosynthates. L2 began to unfold around 8 days after emergence (DAE), rapidly expanded thereafter, and reached a size comparable to the first trifoliate leaf (L1) by 11 DAE, a developmental stage at which L1 is likely to function as a source.

After supplying carbon-14-labeled carbon dioxide (¹⁴CO₂) to tissues below the L2 [L1, primary leaves (PLs), cotyledons, and the surrounding stem], newly synthesized ¹⁴C-photosynthates were visualized by autoradiography (Fig. 1a). Based on autoradiographic patterns, the physiological states of L2 were classified (Fig. 1b). Sink leaves exhibited strong and uniform ¹⁴C accumulation throughout the leaf blade and petiole. Source leaves showed little to no detectable ¹⁴C intensity in both the leaf blade and petiole. Transition leaves displayed an intermediate pattern, characterized by clear ¹⁴C accumulation in the petiole and basal regions but diminished intensity toward the leaf tip. These spatial signatures indicate a tip-to-base maturation gradient during the sink-to-source transition. Autoradiography also revealed that ¹⁴C-photosynthate movement from the main stem into the petiole ceased once L2 became a source, suggesting that the nodal region serves as a regulatory site for transport activity. This observation motivated the inclusion of node tissue together with the leaf tissue for RNA-seq analysis (Fig. 2b).

Figure 1.

Classification of the sink–source status of the L2 based on ¹⁴C distribution and leaf growth. (a) Autoradiographs (left) and photographs (right) of whole aboveground parts for plants in which L2 was classified as a sink, transition, or source. Asterisks indicate leaves supplied with ¹⁴CO₂. Dotted circles indicate the L2. (b) Enlarged views of the L2 blade and the corresponding stem segment containing the L2 node and petiole for each state. The square in the autoradiograph of the leaf indicates the region shown at four-fold magnification to the right. Arrows indicate the excision point of L2 in the petiole. Similar patterns were observed in 7–12 independent plants. (c) Relationship between L2 area and the relative amount of ¹⁴C in L2. The relative ¹⁴C amount represents the proportion of ¹⁴C in L2 relative to the total ¹⁴C detected in all unfolded leaves. Data points are color-coded according to sink, transition, and source states. (d) Relationship between leaf length and leaf area, showing a strong positive correlation (R = 0.94). (e) Relationship between L2 length and the L2/L1 length ratio. Boxes indicate the threshold ranges used to define the sink, transition, and source states. A total of 29 plants were analyzed, including 12 sink, 7 transition, and 10 source plants. (f, g) Autoradiographs to confirm the state of L2. Asterisks indicate leaves that absorbed ¹⁴CO₂. Dotted circles indicate the L2 that absorbed ¹⁴CO₂ when L2 was in the transition (f) and source (g) states. Similar results were obtained from three biological replicates for each. (h, i) Autoradiographs showing the state of the L1 when L2 were defined as a sink. ¹⁴CO₂ was supplied to below L1 (h) or L1 (i). Asterisks indicate leaves that absorbed ¹⁴CO₂. Dotted circles indicate the L1. Similar results were obtained from three biological replicates for each. Scale bars: 5 cm (a, f–i); 1 cm (b). PL, primary leaf; L1, first trifoliate leaf; L2, second trifoliate leaf; L3, third trifoliate leaf.

Figure 2.

Overview of RNA-seq samples and transcriptomic variation among developmental states. (a) Sampling timeline with DAE and mean L2 leaf length (cm) indicated. L2 sink and L1 source were obtained from the same individuals. (b) Schematic illustration of the sampling positions for RNA-seq. (c) PCA of the whole DEG set in leaf tissue. (d) PCA of the whole DEG set in node tissue. Colors indicate the leaf states: L2 sink, L2 transition, L2 source, and L1 source.

To quantify the timing of this transition, the relative ¹⁴C amount of L2 among all unfolded leaves [PLs, L1, L2, and occasionally the third trifoliate leaf (L3)] was calculated (Fig. 1c). The relative ¹⁴C amount of L2 increased with leaf expansion during the sink state but decreased thereafter. It peaked when L2 reached ~18 cm², which corresponded to the boundary between the sink and transition states. Because RNA-seq sampling required classification without ¹⁴C experiments, we used terminal leaflet length and the L2/L1 length ratio as proxies, based on the strong correlation between leaf area and leaflet length (Fig. 1d). A terminal leaflet length of 2.0–3.5 cm (inclusive) defined sink leaves; a length > 3.5 and ≤5.0 cm, together with an L2/L1 ratio ≤ 0.75, defined transition leaves; and >5.0 cm defined source leaves (Fig. 1e). Plants not meeting these criteria were excluded.

The functional relevance of this classification was confirmed by export assays (Fig. 1f and g): L2 did not export assimilates during the transition state, whereas export was evident in source leaves. During the L2 sink state, L1 already acted as a source, exporting assimilates while receiving none from lower leaves (Fig. 1h and i). These L1 leaves were designated “L1 source” for comparison with the L2 source.

Definition of a core sink–source differentially expressed gene set during the L2 sink-to-source transition

Based on the sink-source classification criteria described above, RNA-seq was performed on four defined states: L2 sink, L2 transition, L2 source, and L1 source. Samples were collected between 9 and 11 DAE (Fig. 2a, Supplementary Fig. S1a, Supplementary Table S1). To capture transcriptional changes associated with the sink-to-source transition, RNA was extracted from two tissue types: the terminal leaflet region around the main vein (leaf) and the nodal region including the base of the petiole (node) (Fig. 2b). We included the node tissue because autoradiography showed that ¹⁴C-photosynthate movement from the main stem into the petiole stopped once L2 becomes a source, indicating that the node reflects regulatory changes in transport activity during the transition.

Sequencing generated 6.6–9.5 million filtered reads per sample, of which ~83.5% were successfully mapped to the soybean genome (Supplementary Table S2). Among the 52 872 genes in the reference genome, 19 906 were expressed in leaf tissue, and 22 763 in node tissue (Supplementary Tables S3 and S4). A large proportion of expressed genes were shared between tissues, with 18 057 genes detected in both leaf and node samples (Supplementary Fig. S2a). Differential expression analysis across all four groups (L2 sink, L2 transition, L2 source, and L1 source) identified 2753 differentially expressed genes (DEGs) in leaf tissue and 2358 DEGs in node tissue, defining the whole DEG sets (Supplementary Fig. S1b, Supplementary Tables S5 and S6). In contrast to the broadly shared expressed gene set, the overlap between leaf and node tissues was substantially smaller in the whole DEG sets, with 873 DEGs shared between tissues (Supplementary Fig. S2b).

Principal component analysis (PCA) of the whole DEG sets indicated that principal component 1 (PC1) explained 79.1% and 72.2% of the variance in leaf and node tissues, respectively (Fig. 2c and d). PC1 scores were strongly correlated with leaf length (r = 0.947 for leaf and r = 0.972 for node; Supplementary Fig. S2c and d), demonstrating that developmental progression is the major determinant of transcriptomic variation in both tissues. Because the overlap between leaf and node tissues was substantially smaller in the whole DEG sets than in the expressed gene set (Supplementary Fig. S2a and b), an additional PCA was performed using the combined whole DEG set from leaf and node tissues to further compare transcriptomic variation associated with tissue identity and developmental progression (Supplementary Fig. S2e). In this analysis, PC1 primarily separated samples by tissue type, whereas principal component 2 (PC2) reflected developmental progression along the sink-to-source transition (Supplementary Fig. S2f), with PC2 scores strongly correlated with leaf length (r = 0.934). These results indicate that developmental progression dominates variation within each tissue, whereas tissue identity represents the primary driver of variation across tissues.

To focus on transcriptional changes associated with the sink-to-source transition within the L2 developmental stage, we first refined the L2 DEG set by selecting genes that differed significantly among L2 sink, L2 transition, and L2 source samples, yielding 2284 genes in leaf and 2102 genes in node tissues. To isolate transcriptional changes specifically associated with the sink-to-source transition, genes showing age-dependent expression patterns were excluded based on comparison with L1 source samples. Specifically, genes whose expression in L1 source samples was more similar to L2 sink than to L2 source were regarded as age-dependent and removed from the sink–source DEG set (78 genes in leaf; 102 genes in node; Supplementary methods, Supplementary Figs. S1b and S3). The remaining genes were assigned to five temporal patterns (early, continuous, overall, late, and transient) based on Tukey’s test across the three L2 states, resulting in a sink–source DEG set of 2180 genes in leaf tissue and 1998 genes in node tissue (Supplementary Fig. S1b and c, Supplementary Fig. S3, Supplementary Tables S5–S8, Table 1).

Table 1.

Classification of the sink–source DEG set based on Tukey’s test and expression pattern.

	Type	Grouping (Tukey)	Interpretation	Gene number
				Leaf	Node
Upregulated	Early	a, b, b	Sink < transition = source	106	140
	Continuous	a, b, c	Sink < transition < source	463	748
	Overall	a, ab, b	Sink < source; transition intermediate	13	8
	Late	a, a, b	Sink = transition < source	202	186
	Transient	a, b, a a, b, c	Sink < transition > source; sink = source sink < transition > source; sink > source or sink < source	10	1
Downregulated	Early	a, b, b	Sink > transition = source	117	33
	Continuous	a, b, c	Sink > transition > source	685	804
	Overall	a, ab, b	Sink > source; transition intermediate	73	1
	Late	a, a, b	Sink = transition > source	510	70
	Transient	a, b, a a, b, c	Sink > transition < source; sink = source sink > transition < source; sink > source or sink < source	1	7

Gene ontology enrichment analysis of temporal expression patterns across sink–source states

To associate temporal expression dynamics with biological processes (BP), we performed gene ontology (GO) enrichment analysis (BP terms) for the four major temporal patterns (early, continuous, overall, and late) of the sink–source DEG set (Table 1, Fig. 3a–d; Supplementary Fig. S4a–c; Supplementary Tables S9–S12). Because the transient pattern contained only a small number of genes, it was not included in the GO enrichment analysis; gene information is provided in Supplementary Table S13.

Figure 3.

Pattern-specific GO term enrichment among sink–source DEGs in leaf tissue. (a) Schematic representation of expression patterns within the sink–source DEG set. Pattern definitions are based on Tukey’s test, and are summarized in Table 1. (b, c) Emapplots of enriched GO terms for sink–source DEGs showing (b) increasing and (c) decreasing expression along L2 development, classified into early, continuous, and late patterns. Colors are consistent across panels (a–c) and indicate the early, continuous, and late expression patterns.

No significant GO terms were detected for the overall pattern in either tissue or regulatory direction. In contrast, the early, continuous, and late patterns exhibited distinct enrichments. In leaf tissue, early upregulated genes were specifically enriched in cytokinin signaling, whereas continuous upregulated genes were enriched in sugar metabolism and responses to auxin. Among downregulated genes in leaf tissue, late genes were specifically enriched in chloroplast-related processes, suggesting reduced chloroplast biogenesis during development. In node tissue, the early group yielded few enriched GO terms. Continuous upregulated genes were enriched in transmembrane transport, whereas late upregulated genes were enriched in cell wall-related processes. Downregulated genes in node tissue also showed late enrichment for cell wall-related terms when representative child terms were considered. Together, these results indicate tissue-specific and temporally coordinated regulatory programs during the sink-to-source transition.

Early activation of cytokinin signaling marks the onset of sink-to-source transcriptional reprogramming

Motivated by the specific enrichment of cytokinin-related GO terms in the early upregulated genes of leaf tissue, we first focused on transcriptional changes in cytokinin-related pathways. Candidate genes were curated from previous studies and the soybean cytokinin signaling pathway in Kyoto Encyclopedia of Genes and Genomes (KEGG) (Mens et al. 2018, Nguyen et al. 2021; KEGG PATHWAY Database, https://www.kegg.jp/kegg/pathway.html, gmx04075; Fig. 4a and Supplementary Table S14). Among the curated genes, isopentenyl transferase and cytochrome P450 monooxygenase 735A showed no detectable expression in our dataset. Within the sink–source DEG set, only one Lonely Guy gene, one B-type ARR, and three A-type ARR genes were identified, and all showed upregulation across the sink-to-source progression. Notably, A-type ARR genes constituted a relatively large fraction of expressed cytokinin-related genes and accounted for most cytokinin-related DEGs.

Figure 4.

Differential expression of cytokinin-related genes in leaf tissue. (a) Schematic representation of cytokinin metabolism and signaling pathways, highlighting genes included in the sink–source DEG set. Arrows and blunt-ended lines indicate positive and negative regulatory relationships within the cytokinin signaling cascade, respectively. Heatmaps show Z-scaled expression values for each gene across the L2 sink (si), L2 transition (tr), and L2 source (so) states. Numbers in parentheses indicate the number of genes identified as sink–source DEGs relative to the total number of expressed genes in each category. (b) Expression profiles of cytokinin-responsive downstream genes involved in cell proliferation, cell expansion, and chloroplast maintenance. Numbers in parentheses indicate the number of sink–source DEGs relative to the total number of expressed genes in each functional category. CK, cytokinin; IPT, isopentenyl transferase; CYP735A, cytochrome P450 monooxygenase 735A; LOG, Lonely Guy; CKX, cytokinin oxidase/dehydrogenase; HK, histidine kinase receptor; AHP, histidine phosphotransfer protein; A-ARR and B-ARR, type-A and type-B response regulators; CycD3, cyclin D3; CDK, cyclin-dependent kinase; ANT, aintegumenta; EXPA3, expansin A3; CCS52A, switch 52A; CRF6, cytokinin response factor 6; HEMA1, glutamyl-tRNA reductase 1; GUN4, GUN5, genomes uncoupled 4 and 5; CHLM, magnesium-protoporphyrin IX methyltransferase. Letters indicate significant differences (Tukey’s test, P < 0.05).

We next examined cytokinin-responsive downstream genes associated with cell proliferation, cell expansion, and chloroplast maintenance/suppression of senescence (Fig. 4b and Supplementary Table S14). Among these candidates, cyclin D3, cyclin-dependent kinase, and magnesium-protoporphyrin IX methyltransferase were included in the sink–source DEG set, and all were downregulated during development, consistent with reduced cytokinin-associated promotion of cell division and chloroplast maintenance as L2 acquires source function.

Progressive induction of sugar transporters underlies establishment of source function

As the sink-to-source transition involves a shift from importing to exporting photosynthates, we next examined transcriptional changes in sugar transporters. Similar to many other plant species, soybean primarily transports photosynthate in the form of sucrose (Vernon and Aronoff 1952, Fisher 1970). Based on the current model of sucrose transport (Fig. 5a), we examined the expression profiles of sucrose and monosaccharide transporters in leaf and node tissues (Fig. 5b).

Figure 5.

Expression of sucrose and hexose transporters. (a) Schematic overview of apoplastic phloem loading and unloading based on the current model. In source tissues, sucrose is exported into the apoplast via SWEET and imported into the phloem by SUT. In sink tissues, sucrose is released via SWEET and either taken up directly by SUT or hydrolyzed by CWINV into hexoses, which are subsequently imported by STP (b) Heatmap showing the expression levels of sucrose and hexose transporter genes across samples. The color scale represents log₂ CPM values. Gray boxes indicate genes with no detectable expression. Asterisks on the right indicate genes included in the sink-source DEG set. Letters indicate significant differences (Tukey’s test, P < 0.05).

We surveyed soybean SWEET, SUT, and sugar transport proteins (STP) gene families using curated gene lists (Patil et al. 2015, Guo et al. 2024). In our dataset, expressed SWEET genes were restricted to clades I and III. Several transporters were included in the sink–source DEG set. GmSWEET21 and GmSUT1/2/3 were upregulated toward the source state in both leaf and node tissues, whereas GmSWEET11 and GmSTP7 showed source-associated upregulation only in leaf tissue. GmSWEET30 was specifically expressed in node tissue and increased during development, while GmSWEET6 was leaf-specific and showed higher expression in the sink state. In addition, several transporters showed high, stable expression across the sink–source states without significant changes. Notably, most sink–source DEGs within these transporter families exhibited a continuous expression pattern. Overall, these patterns indicate a continuous transcriptional remodeling of sugar transporter genes in both leaf and node tissues during development.

Tissue-specific regulation of sucrose and callose metabolism during source establishment

To characterize metabolic processes associated with sugar transport during the sink-to-source transition, we examined the expression of genes involved in sucrose metabolism (Fig. 6a and b, Supplementary Fig. S5a), using published gene lists (Su et al. 2018, Xu et al. 2019, Shen et al. 2023). Expression patterns were strongly tissue-dependent: INV transcripts were predominantly detected in leaf tissue, whereas SUS transcripts were more prominent in node tissue. Notably, CWINV transcripts were not detected in either tissue. Most sucrose metabolism genes increased toward the source state in both tissues; however, Glyma.17G045800 (SUS I) showed opposite developmental trends, decreasing in leaf tissue while increasing in node tissue. Most sink–source DEGs related to sucrose metabolism exhibited continuous expression patterns across L2 developmental states.

Figure 6.

Expression of sucrose-metabolizing and callose-related enzymes within the sink–source DEG set. (a) Enzymes involved in sucrose synthesis and degradation, presented as a cascade. (b) Heatmap of sink-source DEGs related to sucrose metabolism shown in (a). (c) Illustration of PD and sieve plate pore closure by callose deposition, together with the enzymes and proteins involved. (d) Heatmap of sink-source DEGs related to callose deposition shown in (c). Color scale represents Z-scaled expression values for each gene. Gray boxes indicate genes that were not classified as sink–source DEGs. Comprehensive heatmaps of expressed genes are shown in Supplementary Fig. S5. SPS, sucrose phosphate synthase; SPP, sucrose phosphate phosphatase; INV, invertase; VINV, vacuole invertase; CINV, cytoplasmic invertase; PDBP, plasmodesmata binding protein; PDCB, plasmodesmata callose-binding protein; BG, β-1,3-glucanases. Letters indicate significant differences (Tukey’s test, P < 0.05).

Because symplastic connectivity also contributes to phloem loading and unloading, we next examined the expression of genes involved in callose turnover (Fig. 6c and d, Supplementary Fig. S5b), based on published gene lists (Zaynab et al. 2024). Callose-related genes generally showed higher expression in node tissue, and sink–source DEGs within this module were limited to callose synthase (CalS) genes. Most sink–source DEGs related to callose metabolism exhibited continuous expression patterns in node tissue, whereas the two sink–source DEGs detected in leaf tissue both showed late-stage upregulation.

Overall, these expression patterns highlight distinct metabolic features of leaf and node tissues during development, with node tissue showing prominent expression of SUS and callose-related genes.

Coordinated changes in nitrogen metabolism and amino acid transport during the sink-to-source transition

To test whether nitrogen-related processes are transcriptionally remodeled as carbon dynamics shift during the sink-to-source transition, we examined the expression of genes involved in nitrate uptake/assimilation and amino acid transport in leaf and node tissues (Supplementary Fig. S6). Genes related to nitrate assimilation were differentially expressed mainly in leaf tissue, whereas amino acid transporter genes were predominantly detected as DEGs in node tissue. Consistent with this pattern, “amino acid transport” was enriched among continuously upregulated genes in node tissue (Supplementary Fig. S4b). Together, these results indicate coordinated transcriptional upregulation of nitrogen metabolism and transport genes during development, suggestive of transport-related remodeling with node tissue showing prominent induction of these genes.

Discussion

In this study, we combined ¹⁴C tracer analysis with RNA-seq to precisely define and analyze the leaf sink-to-source transition in soybean. Rather than comparing source leaves and sink organs such as grains or roots, our analysis focused on the developmental transition of leaves and nodes at the same nodal position. This experimental framework enabled detection of gradual transcriptional changes accompanying source establishment. Furthermore, our approach enabled the physiological classification of leaves into sink, transition, and source states based on actual assimilate allocation and export activity, instead of solely on conventional comparisons between immature and mature leaves.

High-resolution definition of the sink-to-source transition by integrating ¹⁴C tracing and transcriptomics

Autoradiographic analysis revealed that the sink-to-source transition of the L2 occurs within a narrow developmental window, during which carbon import declines and export competence emerges. Under our growth conditions, the maximum terminal leaflet length reached ~7.5 cm. Regression analysis indicated that a leaflet length of 5.0 cm corresponds to an estimated leaf area of ~30 cm², approximately half of the final area (~60 cm² at 7.5 cm; Fig. 1d). Thus, leaves classified as sources in this study had reached roughly 50% of their final size. This observation is consistent with classical physiological studies reporting that dicot leaves begin exporting assimilates when about half-grown in terms of area, achieving a net positive carbon balance (Thrower 1962, Turgeon 2006). Autoradiography further indicated that ¹⁴C import from the main stem into the L2 petiole ceased once L2 became a source, implying regulatory changes at the nodal region and motivating the inclusion of node tissue in the transcriptomic analysis.

PCA further showed that the dominant axis of transcriptomic variation was strongly correlated with leaf length (Supplementary Fig. S2c and d), indicating that developmental progression is a major driver of gene expression changes. Additional PCA integrating both leaf and node whole DEG sets further indicated that tissue identity is the major driver of transcriptomic variation, whereas developmental progression correlated with the sink-to-source transition was captured as the secondary axis (Supplementary Fig. S2e and f). Consistent with this result, leaf- and node-associated genes showed distinct functional enrichments (Fig. 3b and c, Supplementary Fig. S4b and c). In addition, node samples formed more discrete clusters than leaf samples (Fig. 2c and d), suggesting that node tissue exhibits a distinct tissue-specific transcriptional program rather than simply reflecting the developmental status of the subtending leaf. To focus on transcriptional regulation specifically associated with the sink-to-source transition, we therefore filtered out age-dominant genes and further refined the remaining genes based on their temporal expression patterns across L2 states. This strategy enabled us to define a core sink–source DEG set that captures transition-specific transcriptional dynamics and links them directly to changes in carbon flow (Supplementary Fig. S1, Supplementary Fig. S3).

Pattern-based expression analysis reveals staged transcriptional reprogramming during sink-to-source transition

Among sink-source DEGs, the continuous pattern accounted for the largest fraction in both leaf and node tissues (Table 1), indicating that gradual transcriptional reprogramming dominates this developmental process. Notably, leaf tissue also exhibited a large number of late downregulated genes, indicating a pronounced transcriptional shift immediately before the acquisition of full source function.

GO enrichment analysis further highlighted that different biological processes are preferentially associated with specific patterns and tissues (Fig. 3, Supplementary Figs. S4 and S7). In leaf tissue, early upregulated genes were enriched for cytokinin signaling, whereas continuous upregulated genes were associated with carbohydrate metabolism and hormone responses. In contrast, node tissue showed prominent enrichment of transport-related processes in the continuous pattern and cell wall-related processes in the late pattern. Together, these results suggest that source establishment involves coordinated yet temporally distinct regulatory programs that differ between tissues.

In addition to these major temporal patterns, a small number of genes showed transient, transition-specific expression. In leaf tissue, two transcription factors, P. sojae-inducible bHLH transcription factor (PIB1) and homeodomain-leucine zipper protein 57 (HDL57), were included (Supplementary Table S13), both of which have been characterized in soybean as stress-responsive genes (pathogen and salt, respectively; Cheng et al. 2018, Ke et al. 2018). In particular, HDL57 is known to be lowly expressed in young leaves (Ke et al. 2018), and its transient induction may serve as a useful molecular marker of the sink-to-source transition.

Early attenuation of cytokinin signaling accompanies leaf source maturation

Cytokinin signaling emerged as a prominent early-regulated pathway during the sink-to-source transition in leaf tissue (Fig. 3b). Among cytokinin-related genes, A-type ARR transcripts constituted a substantial proportion of sink–source DEGs and were upregulated during the early phase of the transition (Fig. 4a). Given that A-type ARRs function as negative feedback regulators of cytokinin signaling, and that cytokinin regulates genes involved in cell proliferation, cell expansion, and suppression of senescence during leaf development (W. Wu et al. 2021, T. Wu et al. 2021, Huang et al. 2022), the early induction of A-type ARR genes may reflect an attenuation of cytokinin responsiveness. Consistent with this possibility, several genes previously reported to respond to cytokinin and associated with cell proliferation and chloroplast maintenance were downregulated during development (Fig. 4b). However, because these downstream genes are not exclusive markers for cytokinin activity, their downregulation may also reflect the broader developmental progression associated with leaf maturation. Additionally, GO enrichment analysis revealed that late downregulated genes in leaf tissue were enriched for chloroplast-related processes (Supplementary Fig. S4c), indicating the suppression of chloroplast biogenesis at later stages of source establishment.

Exogenous cytokinin application has been reported to enhance sink activity (Nakata and Leopold 1967) and PD transport (Horner and Brunkard 2021). Together, these results suggest that the early engagement of cytokinin negative feedback precedes broader metabolic and structural reprogramming, contributing to a gradual rather than abrupt decline in sink capacity.

Sugar transport-related transcriptional changes during the sink-to-source transition

Autoradiographic analyses were used to visualize the transport of photosynthates, which are predominantly in the form of sucrose. We therefore examined the expression of genes related to sucrose transport and metabolism identified in our dataset (Figs. 5 and 6, Supplementary Figs. S5 and S8). Most SUT genes showing significant expression changes were upregulated as leaves progressed through the sink-to-source transition.

Among the four SWEET clades (I–IV), clade III members are primarily associated with sucrose transport, whereas the other clades predominantly mediate monosaccharide transport (Eom et al. 2015, Su et al. 2024). Notably, in leaf tissue, differential expression during the sink-to-source transition was confined to clade III SWEET genes. GmSWEET11 and GmSWEET21, belonging to clade III, showed increasing expression toward the source state. These transporters have been reported to function in sucrose transport from cotyledons to the hypocotyl during early seedling development (Su et al. 2024), and they may play analogous roles in facilitating sucrose export in developing trifoliate leaves. In contrast, GmSWEET6, belonging to clade III, was uniquely downregulated during the sink-to-source transition and was detected exclusively in leaf tissue. GmSWEET6 has been implicated in sugar delivery to sink tissues in fungal-infected roots (Zheng et al. 2024), suggesting that its higher expression in immature leaves may be associated with sucrose phloem unloading, which is attenuated as leaves acquire source function. GmSUT1/2/3 and GmSWEET21 were also upregulated in node tissue during the sink-to-source transition. These patterns suggest that transcriptional regulation of sugar transporters may be closely linked between leaf and node tissues during the sink-to-source transition.

In the current model of apoplastic unloading, STPs are generally associated with hexose uptake in sink tissues following sucrose hydrolysis by CWINV (Fig. 5a). However, CWINV transcripts were not detected in our dataset (Supplementary Fig. S5a), and no STP genes showed sink-associated upregulation during the L2 sink-to-source transition (Fig. 5b). Instead, GmSTP7 exhibited source-associated upregulation specifically in leaf tissue (Fig. 5b, Supplementary Fig. S8). These results suggest that STP function during soybean leaf development may not be restricted to sink activity and may reflect additional roles of STPs during source establishment.

Phloem unloading in immature leaves is thought to occur predominantly via the symplastic pathway, and the termination of sucrose import has been associated with PD closure (Roberts et al. 2001). Consistent with this model, the late-stage upregulation of GmCalS9 and GmCalS18 in leaf tissue is suggestive of transport-related remodeling, which might be associated with callose dynamics during leaf maturation, although definitive evidence of PD closure requires further high-resolution physiological analysis.

The distinct expression patterns of cytokinin signaling components, sugar transporters, and CalS suggest that these processes may be regulated through different temporal programs during the sink-to-source transition. Whether these programs are causally linked or operate largely independently remains an open question that warrants further investigation.

Node tissue as an active interface during source establishment

Although node samples were classified based on the developmental state of the subtending leaf, juvenile stems undergo active radial growth, and node tissue may therefore exhibit mixed sink–source characteristics.

A distinctive feature of node tissue was the late-stage enrichment of cell wall-related genes (Supplementary Tables S11 and S12). Such structural remodeling may influence local transport properties and contribute to the regulation of sucrose flux between leaves and the main stem during the leaf sink-to-source transition.

Stems are increasingly recognized as dynamic organs that allow radial solute leakage and retrieval, rather than functioning solely as passive transport conduits (De Schepper et al. 2013). Consistent with this view, the present study showed that node tissue exhibited strong upregulation of a broad range of transporter genes, including sugar transporters (Fig. 5, Supplementary Figs. S4 and S6). In addition, several SUS and CalS genes were highly expressed in node tissue (Fig. 6, Supplementary Fig. S5). SUS provides UDP-glucose required for callose biosynthesis (Stein and Granot 2019), and in Arabidopsis, SUSIII has been reported to be preferentially expressed in the phloem (Barratt et al. 2009). Furthermore, AtCalS7 plays a critical role in sieve pore formation during sieve tube development (Xie and Hong 2011), and the soybean homologs GmCalS11 and GmCalS21 also showed increased expression in node tissue in our dataset. These observations suggest that callose-related enzymes in node tissue may be associated with broader vascular maturation processes, including sieve element development, rather than functioning exclusively in immediate transport regulation.

Taken together with the autoradiographic and transporter expression data described above, these transcriptomic features suggest that node tissue is involved in regulating both assimilate import and export and functions as an active interface integrating developmental processes and transport-related functions.

In summary, we functionally defined the leaf sink-to-source transition in soybean using ¹⁴C tracer analysis and revealed a staged transcriptional reprogramming involving early attenuation of cytokinin signaling, sustained induction of sugar transporters, and late activation of callose-related processes. Moreover, our results suggest that node tissue serves as an active interface coordinating assimilate import and export during source establishment.

While the present study focuses on transcriptional regulation, future work integrating high-resolution spatial or compartmentalized metabolite profiling, along with translational and enzymatic activities, will be required to fully bridge the gap between gene expression and actual metabolic states during source establishment. The integration of ¹⁴C tracer analysis with transcriptomics provides a versatile framework that may be applicable to other crops, offering a foundation for understanding spatiotemporal nutrient dynamics and for developing strategies to improve crop productivity.

Materials and Methods

Plant material and growth conditions

Seeds of soybean (G. max. cv. Kosuzu) (Sato Masayuki Seed Company) were germinated in vermiculite and transferred to 1/2 Hoagland hydroponic solution. The plants were grown under a 16 h light (280 μmol m⁻² s⁻¹)/8 h dark cycle at 27°C for 8−11 days. The nutrient solution was refreshed weekly. The length of the terminal leaflets of the L1 and L2 were measured before ¹⁴CO₂ exposure or RNA extraction.

Feeding of ¹⁴CO₂ gas and tracing of translocated ¹⁴C-labeled photosynthates

¹⁴CO₂ feeding experiments were performed 2−3 h after the onset of the light period. In standard experiments, ¹⁴CO₂ was supplied to the region below L2, including L1, PLs, cotyledons, and the surrounding stem. To examine the L1 state, the gas was applied either to the region below L1 (PLs, cotyledons, and the surrounding stem) or directly to L1. To evaluate the export capacity of L2, it was applied directly to L2. To prevent photosynthesis in other parts, the remaining aboveground parts were covered with aluminum foil. Then, the soybean plants were placed in an acrylic box into which ¹⁴CO₂ gas was generated by reacting 5 MBq carbon-14-labeled sodium bicarbonate (NaH¹⁴CO₃) solution (37 MBq/mL, 135 μL; NEC086H005MC, PerkinElmer) with 0.1 M lactic acid (150 μL). Plants were exposed to ¹⁴CO₂ for 15 min. After exposure, ¹⁴C-labeled plants were taken out of the box and the aluminum foil covering them was removed. The plants were further cultivated for 22−24 h in the plant growth chamber.

The distribution of the ¹⁴C in the aboveground parts of the plants was visualized by exposing the samples to an imaging plate (BAS IP MS 2040, FUJIFILM) and subsequently scanning it with an image analyzer (Amersham Typhoon, Cytiva). The relative ¹⁴C amount of L2 was quantified using ImageJ (version 1.53t) according to the following formula:

Unfolded leaves included the PLs, L1 and L2, and L3 if it had begun to unfold at the time of sampling.

Sample definition and tissue sampling

In this study, sampling of L2 at the sink, transition, and source states was conducted using the developmental criteria established from the ¹⁴C tracer analysis (Fig. 1). Specifically, the developmental state of L2 at the time of sampling was defined based on terminal leaflet length. L2 was classified as sink when the length was 2.0–3.5 cm, and as source when it was >5.0 cm. The transition state was defined as >3.5 cm but ≤5.0 cm, only when the L2/L1 length ratio was ≤0.75. The L1 source samples were collected from the same individuals used for the L2 sink sampling. These four sample types are hereafter referred to as “L2 sink,” “L2 transition,” “L2 source,” and “L1 source,” respectively (Supplementary Fig. S1a).

RNA-seq and analysis

Total RNA was extracted from two regions of a single target leaf: a 2–4 mm-wide segment centered on the main vein of the terminal leaflet (hereafter referred to as the “leaf” tissue) and the nodal region at the base of the corresponding petiole, including a short portion of the petiole itself (referred to as the “node” tissue).

All tissue samples were collected 2–3 h after the onset of the light period, immediately frozen in liquid nitrogen, and crushed. Total RNA was extracted using the RNeasy Plant Mini Kit (QIAGEN). A total of three to eight biological replicates were obtained for each state.

RNA samples were sequenced on an Illumina NovaSeq 6000 platform to generate 150-bp paired-end reads. Raw reads were quality-filtered using fastp, mapped to the soybean reference genome (G. max Wm82.a4.v1) using Bowtie2, and gene expression levels were quantified with RNA-Seq by Expectation Maximization (RSEM). Differential expression analysis was performed using the edgeR and limma-voom pipeline in R. DEGs were identified using an Analysis of Variance (ANOVA)-like test in limma (adj.P.val < 0.01, |log₂fold change (FC)| > 1.2) across four states (L2 sink, L2 transition, L2 source, and L1 source). DEGs with a maximum expression level of log₂ counts per million (CPM) > 5 were retained, resulting in 2753 DEGs in leaf tissue and 2358 DEGs in node tissue. These DEGs were further filtered based on significant differences determined by Tukey’s test and expression patterns across developmental states, and categorized into five expression patterns (early, continuous, overall, late, and transient; Table 1, Supplementary Fig. S1c), from which 2180 genes in leaf tissue and 1998 genes in node tissue were defined as the sink–source DEG set. GO enrichment analysis was conducted using the clusterProfiler package with GO annotations obtained from AnnotationHub. Detailed procedures for RNA-seq data processing and DEG classification, and downstream analyses are described in Supplementary methods and Supplementary Fig. S1.

Author Contributions

T.N. and T.K. conceived and designed the study. T.N. and T.K. acquired funding for the project. A.K.-S. performed the experimental work, analyzed the data, and wrote the manuscript. R.S. performed preliminary experimental work. Y.K. assisted with data analysis. Y.K., N.K. and T.K. revised the manuscript. All authors read and approved the final manuscript.

Conflict of Interest

None declared.

Funding

This work was supported by JSPS KAKENHI grant numbers JP20H00437 and JP24H00499, and F-REI (JPFR25040101).

Data Availability

The raw sequencing data for this study have been deposited in the European Nucleotide Archive at EMBL-EBI under accession number PRJEB107831. The data are currently under embargo and will be released upon publication of this article.