Nitrogen Status Rewires Transcriptional Regulation of Dhurrin, a Dual‐Purpose Defense Metabolite in Sorghum bicolor

ABSTRACT

Dhurrin, a cyanogenic glucoside, plays an important role in Sorghum bicolor physiology and defense. The concentration of dhurrin in sorghum is influenced by both nitrogen status and stage of plant organ development. While nitrogen resupply activates the expression of genes for dhurrin biosynthesis, the molecular mechanisms underlying this regulation remain unclear. In this study, we investigated the transcriptional response of sorghum to nitrogen resupply following growth under nitrogen‐limiting conditions. Using a time‐course design, we measured hydrogen cyanide potential (HCNp), growth, and nitrate content at 0‐, 2‐, 6‐, 12‐, 24‐, 36‐, 48‐, and 60‐h after resupply and collected tissue for RNAseq analysis in parallel for analysis of gene expression and construction of gene regulatory networks (GRNs). HCNp (mg g⁻¹ DW) increased significantly in leaf and stem tissues following nitrogen resupply, with increases in the leaf partially driven by continued declines in controls under ongoing nitrogen stress. Expression of the dhurrin pathway genes was upregulated in leaves from 24 h after nitrogen resupply, with diel expression patterns observable over the remaining time points. No upregulation was observed in roots or stems, suggesting that developmental context overrides environmental cues. GRN analysis identified candidate transcription factors regulating dhurrin biosynthesis genes, including members of the MYB, bZIP, and GARP‐type transcription factor families. Some of these candidate transcription factors may be involved in relieving senescence‐associated suppression of dhurrin biosynthesis and link nitrogen signaling to pathway activation. These findings provide new insight into the nitrogen‐responsive regulation of dhurrin in sorghum, highlighting candidate regulators for future functional characterization.

1. Introduction

Specialized metabolism refers to the synthesis and turnover of numerous low‐molecular weight compounds vital for plant success due to their functions in the adaptation of plants to diverse environments, deterring herbivores and pathogens, attracting pollinators, and mitigating harmful effects of abiotic stress (Erb and Reymond 2019; Kessler et al. 2023). In addition to their value in supporting optimal plant growth, specialized metabolites have important industrial uses (e.g., pharmaceutical, food and beverage, bioenergy). The biochemical pathways for production of many specialized metabolites are well‐characterized. These pathways commonly require multiple enzymes to catalyze sequential reactions, often diverging into branches that lead to the synthesis of distinct metabolic products (Ji et al. 2024). Coordinated synthesis and assembly of the biosynthetic machinery are critical for efficient production of specialized metabolites at the right time and place while preventing accumulation of toxic pathway intermediates (Tattersall et al. 2001; Lacchini and Goossens 2020). Various mechanisms have been implicated in the regulation of specialized metabolic pathways, with reports of regulation at the transcriptional, epigenetic, and post‐transcriptional levels (Lacchini and Goossens 2020; Conneely et al. 2022; Li et al. 2024), although a detailed understanding of the molecular regulation for many pathways is currently lacking (Kliebenstein 2023; Ji et al. 2024).

The regulation of specialized metabolism is coordinated within the broader metabolic and developmental context of the plant (Angelovici and Kliebenstein 2022; Kliebenstein 2023). This is important to ensure optimal allocation of a plant's limited resources, such as nutrients, reducing equivalents, and photoassimilates, to either growth or production of specialized metabolites (Gershenzon 1993). Synthesis of chemical defenses against herbivores and pathogens provides an example of this trade‐off. Phytoalexins are only produced upon detection of attack, such as pathogen‐specific molecules or leaf wounding signals (Piasecka et al. 2015). Induction of many phytoalexins is partly mediated by jasmonate signaling that is activated during immune responses (Zhou et al. 2022). Interestingly, as part of the immune response, the jasmonate pathway downregulates growth via the JAZ‐MYC regulatory complex (Campos et al. 2016; Major et al. 2017), providing evidence that the growth‐defense trade‐off is a programmed strategy for optimizing resource use, not just a passive effect of resource depletion (Monson et al. 2022).

Given that specialized metabolites use substrates derived from primary metabolism, such as amino acids, sugars, and organic acids, their synthesis must be coordinated with availability of these metabolites. In contrast to phytoalexins, phytoanticipins are constitutively produced defenses and are often assumed to be less resource‐demanding and thus “cheap” to synthesize (e.g., cyanogenic glucosides) (Piasecka et al. 2015), although this is not always the case—especially when resources are limiting. Plant nutrient status is communicated via signaling from nutrient availability sensors, such as NITRATE TRANSPORTER 1.1 (NRT1.1; nitrate), PHOSPHATE STARVATION RESPONSE 1 (PHR1; phosphate), and hexokinase 1 (HXK1; glucose) (Ho et al. 2009; Pant et al. 2015; Liu et al. 2017; Li et al. 2021). These cues are integrated into metabolic signaling pathways that broadly coordinate growth with resource availability. The master metabolic regulators, target of rapamycin (TOR) and SNF1‐related protein kinase (SnRK), promote and antagonize growth, respectively, by influencing hormone signaling pathways and transcriptional control (Baena‐Gonzalez and Lunn 2020; O'Leary and Plaxton 2020; Li et al. 2021). Under nutrient limiting conditions, the SnRK signaling pathway antagonizes TOR signaling and suppresses growth while activating catabolic pathways (Margalha et al. 2019). Among large‐scale changes in plant metabolism, switching between TOR and SnRK signaling results in altered metabolic profiles. Altered nutrient status is known to influence accumulation of specialized metabolites (Fritz et al. 2006; Rubin et al. 2009; Pant et al. 2015), and the JAZ‐MYC complex has been implicated in the coordination of tryptophan supply with glucosinolate biosynthesis in Arabidopsis thaliana (A. thaliana) (Guo et al. 2022). However, for most specialized metabolites, the mechanism and regulatory components controlling biosynthesis in the context of nutrient supply have not yet been elucidated.

Sorghum ( Sorghum bicolor [S. bicolor] L. Moench) produces dhurrin, a nitrogen‐containing defense metabolite classed as a cyanogenic glucoside (Gleadow and Møller 2014). Dhurrin is considered a phytoanticipin and has a role in herbivore defense. While dhurrin is a stable compound, hydrolysis of its glucosidic bond releases toxic hydrogen cyanide (HCN), catalyzed by endogenous β‐glucosidases that are spatially separated from their substrate but come into contact during tissue disruption (e.g., chewing), referred to as cyanogenesis or bioactivation (Gleadow and Møller 2014). Synthesis of dhurrin in sorghum is the highest in juvenile stages of growth (Miller et al. 2014), and both content and concentration can be extremely high in young tissues such as coleoptiles (Halkier and Møller 1989). However, dhurrin content is also influenced by environmental conditions including soil nitrogen availability (Hayes et al. 2015). Moreover, dhurrin is turned over within the plant (Bough and Gander 1971; Adewusi 1990). There is strong evidence that dhurrin is turned over without the release of HCN by an endogenous recycling pathway that enables the recovery of the nitrogen, which can be channeled back into primary metabolism (Pičmanovà et al. 2015; Nielsen et al. 2016; Bjarnholt et al. 2018; Sánchez‐Pérez and Neilson 2024). Nitrogen can also be recovered from the products released during bioactivation as part of the HCN detoxification pathway (Jenrich et al. 2007). Due to the ability of plants to recycle nitrogen in these ways, it has been proposed that dhurrin may serve dual functions, in defense and as a nutrient store (Møller 2010; Gleadow and Møller 2014; Bjarnholt et al. 2018). Similar roles for cyanogenic glucosides in other species have also been proposed (Selmar et al. 1988; Siegień et al. 2021), consistent with the hypothesis that the cost of production of dhurrin may be offset by the ability to store and recycle nitrogen—and thus may provide a selective advantage (Neilson et al. 2013).

Like other specialized metabolic pathways, much is known about the biochemistry and physiology of dhurrin production. Less is known about how the pathway is regulated. Dhurrin is derived from the aromatic amino acid tyrosine and requires the activity of three structural enzymes for its synthesis: cytochromes P450 SbCYP79A1 and SbCYP71E1 and the UDP‐glycosyltransferase SbUGT85B1 (Koch et al. 1995; Sibbesen et al. 1995; Bak et al. 1998; Jones et al. 1999; Kahn et al. 1999). These enzymes form a metabolon anchored to the endoplasmic reticulum, and along with cytochrome P450 oxidoreductase (SbPOR) for supply of NADPH, that enables channeling of the intermediate oxime and hydroxynitrile compounds to form the stable dhurrin monoglucoside (Laursen et al. 2016). There are three POR isoforms in S. bicolor , with SbPOR2b considered to function in dhurrin biosynthesis (Laursen et al. 2016). Apart from SbPOR2b, the enzymes are specific to dhurrin biosynthesis, and while dhurrin derivatives are generated (Selmar et al. 1996; Pičmanovà et al. 2015; Montini et al. 2020), dhurrin is the major compound produced from this relatively simple pathway. The genes encoding the dhurrin biosynthesis enzymes, along with SbMATE2 involved in dhurrin transport (Darbani et al. 2016), are clustered within a 70‐kb region on chromosome 1 (Takos et al. 2011), likely facilitating the co‐expression of these genes (Rosati et al. 2019; Nützmann et al. 2020). The dhurrin biosynthesis pathway is transcriptionally regulated during organ and plant development and can be induced by methyl jasmonate as part of the wounding response (Sohail et al. 2023). A GATA‐family transcription factor (TF) SbGATA22 has a role in regulating expression of genes involved in dhurrin biosynthesis during development (Rosati et al. 2024). The dhurrin biosynthesis genes SbCYP79A1 and SbCYP71E1 are induced by nitrate application (Busk and Møller 2002; Rosati et al. 2024), and nitrogen availability alters expression of genes in the pathway (Blomstedt et al. 2018; English et al. 2025), with greater accumulation of dhurrin under high‐nitrogen regimes (Myrans et al. 2021). However, no TFs or regulatory mechanisms have been implicated in the nitrogen‐responsive control of dhurrin biosynthesis or, indeed, recycling.

This study aimed to investigate the regulatory mechanisms underlying control of dhurrin accumulation in response to a high nitrogen treatment, with a focus on identification of candidate TFs that mediate increased transcription of the dhurrin regulon. Gene coexpression networks and gene regulatory networks (GRNs) have previously been used to analyze complex regulatory interactions controlling biosynthesis of specialized metabolites including alkaloids (Sun et al. 2020; Hong et al. 2025), glucosinolates (Wisecaver et al. 2017), phenolics (Yang et al. 2017), and cyanogenic glucosides (Zhang et al. 2024; Katamreddy et al. 2025). Regulatory network analysis has also been used extensively to characterize the dynamic response of plants to changes in nitrogen availability over time (Varala et al. 2018; Du et al. 2020), revealing dynamic transcriptional responses coordinated by nitrogen signaling networks (Shanks et al. 2024). As nitrogen application is known to induce expression of the dhurrin biosynthesis pathway, a nitrogen surge experiment was designed where a high nitrogen treatment was used to induce nitrogen‐responsive transcriptional networks, and we hypothesized that this would include networks involved in upregulation of dhurrin biosynthesis. Changes in the transcriptional and chemical response were analyzed from eight time points across a 60‐h sampling period, enabling the investigation of the nitrogen response over time. To maximize detection of changes in dhurrin related to the nitrogen response, the plants were grown beyond early development to reduce dhurrin accumulation (Miller et al. 2014). Plants were initially grown under nitrogen‐limiting (0.5 mM N) conditions until N‐treatments (12 mM or 0 mM N) were imposed at the 6‐leaf stage. Tissues were sampled from leaf, stem, and root tips.

2. Results

2.1. Impact of Nitrogen Treatment on Growth Not Detected Within the Experimental Window

The plants in this experiment were grown with limited nitrogen (0.5 mM N) until they reached the 6‐leaf stage (55 days). At this point, half the plants were treated with an increase of nitrogen (12 mM N), while the other half received no additional nitrogen (control, 0 mM N) (Figure S1). Application of the 12‐mM nitrogen treatment had little impact on the growth of plants for 60 h posttreatment, with no visible differences in plant size over the duration of the sampling period (Figure S2). No significant differences were detected in leaf number, height, or biomass over this period, in either treated plants or in controls that did not receive the nitrogen treatment (Figure S3a–c). Also, no significant changes in the root:shoot ratio of nitrogen‐treated or control plants were detected (Figure S3d).

2.2. Nitrogen Treatment Increased Nitrate Concentration in Roots and Sheath but Not Leaves

Nitrate concentration was determined for leaf, sheath, and root tissue samples from plant tissue types that were separated prior to grinding. Leaf nitrate was similar at all time points, whether or not plants had been subject to the nitrogen treatment, with an overall mean of 1.32 ± 0.09 mg g⁻¹ nitrate in controls and 1.60 ± 0.07 mg g⁻¹ nitrate for nitrogen‐treated plants (pooling all data) (Figure 1a). There was, however, a significant difference in nitrate concentration in the sheath tissue both with treatment (F(1, 75) = 676.9, p < 0.0001) and time point (F(7, 75) = 76.45). The treatment effect could be seen as early as 12‐h post nitrogen resupply in sheath tissue (p < 0.0001), with approximately double the nitrate concentration in the nitrogen‐treated plants (1.38 ± 0.07 mg g⁻¹ nitrate) compared to nitrate in the sheaths of control plants (0.61 ± 0.04 mg g⁻¹) (Figure 1b). At the final time point, nitrate in samples from nitrogen‐treated plants reached 6.13 mg g⁻¹ but remained at an average of 0.60 mg g⁻¹ in control plants.

FIGURE 1.

Chemical composition of Sorghum bicolor plants harvested at intervals over 60 h after delivery of a 12 mM N‐surge treatment or a 0 mM N of control treatment. Nitrate concentration of samples for (a) leaf (n = 6), (b) sheath (n = 6), and (c) roots, (n = 6). Chlorophyll concentration (d) of the 6^th fully expanded leaf of plants in response to nitrogen surge treatment, as determined using a SPAD meter (n = 9). Measurements were taken prior to, and at five time points within the experimental time frame. Hydrogen cyanide potential (HCNp) of samples for (e) leaf (n = 6), (f) sheath (n = 6), (g) stem (n = 4), and (h) roots (n = 6). Plants were grown at 0.5 mM N until the 6–7 leaf stage prior to treatment. Columns represent mean ±SE. Data were analyzed using a two‐way ANOVA and Bonferroni's test for multiple comparisons. Significant differences between means at each time point is indicated by the following: * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001.

The nitrogen treatment resulted in a significant increase in nitrate concentration of root tissue (Figure 1c) (F(1, 72) = 295.9, p < 0.0001) and was detected from as early as 2‐h post nitrogen application (p < 0.0001). The highest nitrate concentration was observed in nitrogen‐treated plants at the 36‐h time point, with 9.72 ± 2.06 mg g⁻¹ nitrate. Root nitrate then decreased again across the final time points, with 6.74 ± 1.73 mg g⁻¹ nitrate recorded at 60‐h post nitrogen application, approximately 10 times higher than the 0.69 ± 0.21 mg g⁻¹ nitrate recorded for controls at the same time point.

2.3. Nitrogen Treatment Increased Leaf Chlorophyll

Chlorophyll concentration was measured with a SPAD meter (μmol m⁻²) on the sixth fully expanded leaf of plants at the time of harvest. Chlorophyll concentration of nitrogen‐treated plants increased over the duration of the sampling period (F(1,87) = 8.124, p = 0.0055) (Figure 1d). Concentrations for nitrogen‐treated plants were statistically higher at the 36‐h (p = 0.0391) and 60‐h time points (p < 0.0001), with concentrations 25 and 50 μmol m⁻² higher, respectively, than control plants.

2.4. Hydrogen Cyanide Potential (HCNp) Increased in Shoots With Nitrogen Application but Not in Roots

HCNp is the total amount of HCN capable of being released if all the dhurrin in a sample were hydrolysed. As HCN produced by other processes in the plant is negligible (e.g., ethylene production), HCNp is used as a proxy for dhurrin and is comparable to methods where dhurrin is quantified directly (Gleadow et al. 2016). HCNp was determined for leaf, sheath, stem, and root tissue (Figure 1e–h). At the baseline harvest, leaf HCNp was slightly higher than sheath and root, ranging from 0.11 to 0.20 mg g⁻¹ DW (Figure 1e). In contrast, the baseline stem HCNp was ~10 times higher than leaf, sheath, or root tissue (Figure 1g). HCNp was significantly higher in the nitrogen treated leaf (F (1, 75) = 24.39, p < 0.0001) and sheath samples (F(1, 74) = 7.757, p = 0.0068) (Figure 1e,f). However, significant interactions between treatment and time point for leaf (F(7, 75) = 4.391, p = 0.0004) and sheath (F(7, 74) = 3.087, p = 0.0065), as well as post hoc analyses, indicated that higher HCNp in nitrogen‐treated plants was only detected at evening time points: 36 and 60 h. In leaves, this difference appears to result from a decrease in the HCNp of control plants, as average HCNp values are maintained in nitrogen‐treated plants (0.17 ± 0.02 mg g⁻¹ HCNp at baseline to 0.18 ± 0.02 mg g⁻¹ HCNp at the 60‐h time point), while the values for control plants at the final time point are lower (0.08 ± 0.01 mg g⁻¹ HCNp) (Figure 1e). Likewise, for sheath tissue, post hoc analyses showed that nitrogen‐treated plants had significantly higher HCNp than the controls at 36 and 60 h (Figure 1f). However, in this case, this difference appears to be driven by an increase in HCNp in the nitrogen‐treated plants, rather than a decrease in the controls. As at the final 60‐h time point, nitrogen‐treated plants averaged 0.10 ± 0.02 mg g⁻¹ HCNp compared to 0.02 ± 0.00 mg g⁻¹ HCNp in controls and more than double the baseline value of 0.03 ± 0.01 mg g⁻¹ HCNp for sheath tissue. The highest HCNp was detected in the stem, with average concentrations ranging from 0.69 ± 0.01 to 1.31 ± 0.21 mg g⁻¹ HCNp across the sampling time frame (Figure 1g). The values for stem HCNp are up to 10 times higher than the HCNp values for leaf, sheath, and roots; however, HCNp concentrations varied widely in the stem, and no significant differences were detected between nitrogen‐treated and control plants in this experiment. Similarly, the nitrogen resupply did not increase the HCNp in roots, with no significant differences detected between nitrogen‐treated plants and controls at any time point (Figure 1h).

2.5. Timing, Magnitude, and Duration of Transcriptomic Response to Nitrogen Treatment Differs According to Plant Organ

The transcriptional response of leaf, stem, and root tips to delivery of nutrient solution containing 12‐mM nitrogen was investigated across seven intervals from 2 to 60 h after treatment application, with comparisons made relative to control plants that received a nutrient solution containing 0 mM of nitrogen (Figure S1a). Principal component analysis (PCA) showed strong grouping of samples by organ (Figure S4d). Further analysis of samples within organ type showed that in the leaf, clustering along PC1 was driven by time of day (AM and PM), while clustering along PC2 was driven by differences in treatment group (Figure 2a). In stems, clustering patterns were not as clear as in the leaf, with differences in clustering along PC1 driven by treatment group and differences in time along PC2 (Figure 2b). For root tips, a strong effect of treatment was observed (Figure 2c), with control and nitrogen‐treated samples diverging along PC1. However, there was less influence of time of day.

FIGURE 2.

Global changes in transcriptome in response to nitrogen resupply. Principal component analysis (PCA) of RNAseq samples performed separately for each tissue type, (a) leaf, (b) stem, and (c) root tips. PCA was performed on log2‐transformed counts for all samples within tissue type after filtering to remove low expressed genes and TMM normalization. PC1 and PC2 represent the top two components of variation. Treatment represented by shape (control—circles; nitrogen—triangles). Time point indicates the number of hours after treatment, with shade scaled from light to dark with increasing time. (d) Counts of significant (FDR < 0.05) transcript‐level DE genes upregulated or downregulated in nitrogen‐treated plants compared to controls. Transcripts were filtered by TPM ≥ 5, FC ≥ 5 (upregulated genes) or FC ≤ −5 (downregulated genes). Heatmaps of z‐score transformed mean log₂₍TPM) for transcript‐level DEGs by tissue type for (e) root tips, (f) leaf, and (g) stem (n = 3, except root 6‐h control, 24‐ and 60‐h nitrogen treatment where n = 2). Hierarchical clustering for rows (transcripts) performed using Euclidean distance with complete linkage method. Dendrogram on left represents transcript clustering, and number represents clustering of transcripts according to k‐means clustering (k = 5). Color scale represents z‐score‐normalized expression levels. Sample groups indicating treatment group are shown in black and gray. Samples are labeled with the number of hours after treatment.

Differential expression analysis in leaves, stems, and root tips showed clear differences in the response to nitrogen resupply over time (Figure S5). In leaf tissue, the number of differentially expressed (DE) transcripts detected increased from 6 h and remained high for the duration of the experiment (Figure S5). In stems, the number of DE transcripts was highest at 12–24 h (Figure S5), while in root tips, numbers increased as early as 2‐h posttreatment and remained elevated up to 12 h (Figure S5).

The greatest change in transcriptional response to nitrogen resupply was observed in leaf tissue, with a cumulative 2693 transcripts upregulated or downregulated across all time points (Figure 2d). Root tips displayed a faster transcriptional response to nitrogen application compared to leaves and stems, with a total of 141 DE genes at 2 h after treatment, compared to 39 genes in stem and only one in leaf tissue at the same time point (Figure 2d). The majority of the DE transcripts in root tips and stems were upregulated at this first time point. The biggest transcriptional response in roots was sustained through the 6‐ and 12‐h period with greater than 200 upregulated and 200 downregulated genes detected at both time points; however, from 24 h onwards, fewer DE transcripts were detected in root tips (Figure 2d). Hierarchical clustering of DE transcripts for root tips shows distinct transcriptional expression profiles between time points in response to the nitrogen treatment (Figure 2e). Cluster 1 is characterized by transcripts that were rapidly upregulated from 2 to 6 h but show a temporal shift as expression decreases at later time points. The DE transcripts in cluster 2 also show a strong but reverse pattern of expression, with low levels at 2–12 h followed by a marked increase in expression after 24 h. Clusters 3, 4, and 5 include transcripts with similar expression patterns across time points within treatments groups, with transcripts in Clusters 3 and 5 largely downregulated in response to nitrogen application, and those in Cluster 4 upregulated (Figure 2e).

Transcriptional responses to nitrogen treatment in leaf and stem were not as rapid as those observed in roots, with greater changes occurring 12–24 h after treatment (Figure 2d). For leaves, 385 DE transcripts were identified as either upregulated or downregulated at 24 h (Figure 2d), and the number of DE transcripts remained elevated across the subsequent time points. For example, at the final 60‐h time point, 615 genes were detected as either upregulated or downregulated in leaf, with approximately double the number of upregulated compared to downregulated genes at the 48‐ and 60 h time points (Figure 2d). In addition to the treatment effects, hierarchical clustering in leaf tissue shows that diel changes in expression were a strong factor in cluster groupings, with Clusters 1, 2, and 3 characterized by transcripts with contrasting circadian expression patterns (Figure 2f).

The stem showed a similar transcriptional response to leaf, where the numbers of DE transcripts were higher at the 24‐h time point than at the earlier time points, with a total of 748 DE transcripts (Figure 2d). Of these 748 DE transcripts, most (594) were downregulated in the stem. In contrast to leaf tissue, fewer DE transcripts were detected over the remaining three time points from 36 to 60 h after treatment, totalling 286 DE transcripts—less than half the number of downregulated transcripts detected at 24 h. In this way, the transcriptional response in stems was more similar to the response observed in roots. While there was some clustering of stem samples according to time‐of‐day indicative of diel expression (Cluster 3), this pattern was much weaker than in the leaf (Figure 2g).

2.6. Gene Ontology (GO) Analysis Reveals Enrichment for Terms Related to Primary Nitrogen Metabolism

Over the entire 60 h of the experiment, there were 54 unique DE transcripts common to all three tissue types (Figure 3a), and these transcripts were significantly enriched for the GO terms related to nitrate responses and nitrogen assimilation (Figure 3b). The largest number of DE transcripts was identified in leaf tissues, with 1524 unique transcripts either upregulated or downregulated across one or more time points across the 60‐h experiment, and of these, 1190 were specific to leaf tissue (Figure 3a). GO enrichment analysis for leaf‐specific DE transcripts indicates they have largely leaf‐specific functions, for example, “photosynthesis” (Figure 3c).

FIGURE 3.

Global changes in transcriptome in response to nitrogen resupply. (a) Comparison of unique and shared transcripts between tissue types that were differentially expressed in at least one time point between nitrogen and control treatments. (b) Gene ontology (GO) enrichment analysis for the 54 common transcripts differentially expressed in all tissues in at least one time point (Biological process, FDR < 0.05). The Arabidopsis thaliana symbols for annotated genes in each GO category are listed under “genes” NiR1 = Nitrite reductase 1; CLC‐A = Chloride channel A; BT1 = BTB and TAZ domain protein 1; GLB1 = hemoglobin 1; MDAR6 = Monodehydroascorbate reductase 6; ROC2 = Rotamase cyclophilin 2; GS2 = Glutamine synthetase 2; GOGAT = glutamine synthase. (c) GO term enrichment for differentially expressed (DE) transcripts specific to tissue type (leaf, stem, root) and DE transcripts detected in all tissues (all) at any time point (biological process, FDR < 0.01). See Figures S6–S8 for detailed information on enriched GO terms by tissue type.

The number of DE genes in roots and stem was approximately equal (1063 and 980, respectively) with 798 and 729 unique to roots and stems, respectively (Figure 3a). Stem tissue had the largest number of significantly enriched GO terms, which included many terms related to biotic stress signaling and response, for example, “regulation of defence response,” “response to jasmonic acid,” and “response to chitin” (Figure 3c). A smaller set of enriched GO terms was detected for DE genes in root tips, including terms related to secondary metabolism, such as “sterol metabolic process” and “secondary metabolism biosynthetic process,” and others related to the generation of reducing power, including “pyridine‐containing compound metabolic process” and “oxidation–reduction process” (Figure 3c).

For leaf at 24 h, GO analysis indicated that genes and pathways related to nitrogen metabolic and biosynthetic processes were significantly upregulated (including terms summarized under “nitrogen compound metabolic process” and “glutamine biosynthetic process”) (Figure S6). In addition, at the 48‐h time point, upregulated DE transcripts in leaf tissue were enriched in terms related to photosynthesis (Figure S6). Several GO terms related to dhurrin biosynthesis were also significantly enriched in the upregulated gene sets for leaf tissue at 36, 48, and 60 h. These categories (e.g., GO:0010132—dhurrin biosynthetic process; GO:0019756—cyanogenic glycoside biosynthetic process) are not present on the GO term plot as only two genes are annotated with this category in the dataset (SbCYP79A1 and SbCYP71E1) and the plot has excluded significant GO terms with counts of annotated genes fewer than five. These genes have been investigated using other methods (a complete list of GO‐term results can be found in Table S4).

In stem, GO enrichment analysis showed that most terms were identified at the 24‐h time point, and these were mainly associated with downregulated processes (Figure S7). A large number of these downregulated processes were related to defense, for example, jasmonic acid biosynthesis and signaling, phytoalexin and phenylpropanoid biosynthesis, and regulation of immune response. Only a few upregulated GO terms were identified, and these included terms related to protein biosynthesis and cellular function (summarized under the categories “ribonucleoprotein complex assembly,” “peptide biosynthetic process,” “cellular homeostasis,” and individual terms GO:0019438—cellular amide metabolic process and GO:0006414—translational elongation).

In contrast to leaf and stem where the most significantly enriched GO‐terms were associated with either upregulation or downregulation, enriched GO terms in root tips were associated with both upregulated and downregulated processes (Figure S8). Significantly enriched GO terms included those related to nitrogen uptake and assimilation (“organonitrogen compound metabolic process,” “nitrogen compound transport”), which were upregulated from 6 h after nitrogen application. GO terms associated with primary metabolic pathways were generally upregulated in roots (pentose‐phosphate shunt, carbohydrate metabolic process, and cellular amino acid metabolic process). GO terms associated with hormone pathways (brassinosteroid biosynthetic process and strigolactone biosynthetic process) were enriched among downregulated genes in roots (Figure S8).

2.7. Expression of Key Nitrogen Genes Induced by Nitrogen Treatment

In nitrogen‐treated plants, upregulation of genes involved in nitrogen uptake, transport, and assimilation was identified in root tips, stem, and leaf tissues (Tables 1, 2, 3). Upregulation of genes in the nitrate assimilation pathway (nitrate reductase (NR), nitrite reductase (NiR), urophorphyrin methyltransferase (UPM1)) occurs as early as 2 h in root (Table 1) and stem (Table 2) tissue but not until 6 h in leaf tissue (Table 3). Genes encoding several isoforms of enzymes involved in the ammonium assimilation pathway (glutamine synthase [GS1], glutamine synthetase 2 [GS2], and NADH‐dependent GS1 [NADH‐GOGAT]) are also upregulated in response to nitrogen application in root tips, stems, and leaves (Tables 1, 2, 3), and expression remained elevated in nitrogen‐treated samples throughout the 60‐h sampling time frame.

TABLE 1.

Nitrogen assimilation pathway genes, root tissue expression in TPM.

Note: TPM values are colored according to the lowest and highest TPM values across each row, scaled using the “scale” function in R.

Abbreviations: FD‐GOGAT, glutamate synthase 1; GDH2, glutamate dehydrogenase 2; GS1, glutamine synthase; GS2, glutamine synthetase 2; NADH‐GOGAT, NADH‐dependent GS1; NiR, nitrite reductase 1; NR, nitrate reductase 1; UPM1, urophorphyrin methylase 1.

TABLE 2.

Nitrogen assimilation pathway genes, stem tissue expression in TPM.

Note: TPM values are colored according to the lowest and highest TPM values across each row, scaled using the “scale” function in R.

Abbreviations: GDH2, glutamate dehydrogenase 2; GS1, glutamine synthase; GS2, glutamine synthetase 2; NADH‐GOGAT, NADH‐dependent GS1; NiR, nitrite reductase 1; NR, nitrate reductase 1; UPM1, urophorphyrin methylase 1.

TABLE 3.

Nitrogen assimilation pathway genes, leaf tissue expression in TPM.

Note: TPM values are colored according to the lowest and highest TPM values across each row, scaled using the “scale” function in R.

Abbreviations: GDH2, glutamate dehydrogenase 2; GS1, glutamine synthase; GS2, glutamine synthetase 2; NADH‐GOGAT, NADH‐dependent GS1; NiR, nitrite reductase 1; NR, nitrate reductase 1; UPM1, urophorphyrin methylase 1.

Construction of a GRN for root tips at the 12‐h time point includes many genes involved in nitrogen‐assimilation (NiR and GS), nitrogen transport (NRT1.1, ammonium transporter 2 member 1 (AMT2;1), amino acid biosynthesis (asparagine synthase [AS], aspartate aminotransferase 1 [ASP1]), and amino acid transport (amino acid permease 7 [AAP7], amino acid transporter AVT6A [AVT6A]) (Figure S9). TFs known to regulate nitrogen responses were present in the 12‐h‐GRN, including Hypersensitive to low Pi‐elicited primary Root Shortening 1 (HRS1), LOB domain‐containing protein 37 (LBD37), and ABRE‐binding factor 4 (ABF4) that are predicted to bind to the promoters of genes involved in nitrogen assimilation and transport (Figure S10). Additional TFs with known roles in regulation of nitrogen responses were DE in root tips (Figure S10) but are not present in the 12‐h GRN as they did not pass the filtering thresholds used for GRN construction. Several of these nitrogen‐related TFs are also DE in stem and leaf at several time points. For example, Sobic.001G078900, an orthologue of the nitrate‐responsive AtHRS1, is DE between nitrogen and control treatments at a minimum of one time point in each of leaf, stem, and root tips (Figures S11 and S12). Many other TFs in the present study with altered expression in response to nitrogen treatment do not have known roles in the regulation of nitrogen metabolism or responses to variation in plant nitrogen status (Figures S11 and S12).

2.8. Nitrogen Treatment Induces Expression of Dhurrin Metabolism Genes in Leaves but Not Stems or Roots

Transcript levels of the dhurrin biosynthesis pathway genes SbCYP79A1, SbCYP71E1, and SbUGT85B1 were very low in leaf tissue of control plants at all time points (Figure 4a). Expression in the leaf was also very low in the nitrogen‐treated plants at time points up to and including 12 h (Figure 4a). In nitrogen‐treated plants at 24 h, transcripts of SbCYP79A1 had a log₂FC of 8.2 (FDR < 0.01), corresponding to ~40‐fold upregulation in comparison to controls. Expression of SbCYP79A1 remained elevated in leaf tissue of nitrogen‐treated plants over the following time points, with ~200‐fold greater expression at 48 h (log₂FC = 15.7, FDR < 0.001) (Figure 4a and Table 4). Transcripts for two other biosynthesis genes, SbCYP71E1 and SbUGT85B1, were also upregulated in leaf tissue of nitrogen‐treated plants after 24 h, with the greatest increases occurring at 48 h where expression increased ~90‐fold (log₂FC = 6.5, FDR < 0.0001) and ~200‐fold (log₂FC = 7.7, FDR < 0.001), respectively, in comparison to controls (Figure 4a and Table 4). In addition to upregulation by nitrogen resupply, transcript levels of all three biosynthesis genes showed diel cycling expression patterns in leaf tissue, with greater transcript levels at morning time points (24 and 48 h) in comparison to evening time points (36 and 60 h) (Figure 4a). A diel expression pattern was also observed for SbPOR2a and SbPOR2b. SbPOR2b is an essential but not specific component of the dhurrin biosynthesis pathway and was not DE between control and nitrogen‐treated plants.

FIGURE 4.

Expression of dhurrin biosynthesis genes in (a) leaf, (b) stem, and (c) root tips across 60 h after delivery of a 12 mM N of solution or a 0 mM N of control for Sorghum bicolor plants grown at 0.5 mM until the 6–7 leaf stage (n = 3, except root 24‐ and 60‐h nitrogen treatment where n = 2). Data for the root 36‐h time point excluded due to insufficient replicates. Asterisk indicates significant differential gene expression between nitrogen and control within a time point (FDR < 0.05). CYP79A1, Cytochrome P450 79A1; CYP71E1, Cytochrome P450 71E1; UGT85B1, UDP‐dependent glycosyltransferase 85B1; POR2a‐c, NADPH‐dependent phytochrome oxidase isoform a–c.

TABLE 4.

Dhurrin metabolsim genes in leaf, expression in TPM.

Note: TPM values are colored according to the lowest and highest TPM values across each row, scaled using the “scale” function in R.

Abbreviations: CAS, cyanoalanine synthase; CGTR, cyanogenic glucoside transporter; CYP, cytochrome P450; DHR, dhurrinase; GST, glutathione transporter; HNL, hydroxynitrile lyase; MATE, multidrug and toxic compound extrusion transporter; NIT, nitrilase; POR, cytochrome P450 oxidoreductase; UGT, UDP‐glucosyltransferase.

All three dhurrin biosynthesis genes were expressed in control and nitrogen‐treated plants at high levels in stems throughout the sampling time frame (Figure 4b and Table 5). Transcript levels for SbCYP79A1 and SbCYP71E1 are higher in stem than leaf, with a mean TPM of 939 for SbCYP79A1 at 60 h for control and a mean TPM of 708 for SbCYP71E1 at the same time point (Table 6). Interestingly, the expression of SbUGT85B1 is lower than SbCYP79A1 and SbCYP71E1 in stem tissue, while in leaf and roots, expression of SbUGT85B1 is similar or higher that SbCYP79A1 and SbCYP71E1 (Tables 4 and 6).

TABLE 5.

Dhurrin metabolsim genes in stem, expression in TPM.

Note: TPM values are colored according to the lowest and highest TPM values across each row, scaled using the “scale” function in R.

Abbreviations: CAS, cyanoalanine synthase; CGTR, cyanogenic glucoside transporter; CYP, cytochrome P450; DHR, dhurrinase; GST, glutathione transporter; HNL, hydroxynitrile lyase; MATE, multidrug and toxic compound extrusion transporter; NIT, nitrilase; POR, cytochromeP450 oxidoreductase; UGT, UDP‐glucosyltransferase.

TABLE 6.

Dhurrin metabolsim genes in root tips, expression in TPM.

Note: TPM values are colored according to the lowest and highest TPM values across each row, scaled using the “scale” function in R.

Abbreviations: CAS, cyanoalanine synthase; CGTR, cyanogenic glucoside transporter; CYP, cytochrome P450; DHR, dhurrinase; GST, glutathione transporter; HNL, hydroxynitrile lyase; MATE, multidrug and toxic compound extrusion transporter; NIT, nitrilase; POR, cytochrome P450 oxidoreductase; UGT, UDP‐glucosyltransferase.

Similar to stems, the genes in the dhurrin biosynthesis pathway were not DE in root tips in response to nitrogen resupply (Figure 4c and Table 6). However, transcript levels were lower than those reported in stem, with maximum mean TPM values of 154 and 174 for SbCYP79A1 and SbCYP71E, respectively, at the 60‐h time point for control plants and even lower for nitrogen‐treated plants. Transcripts for SbUGT85B1 were expressed slightly higher than the cytochromes P450 and were upregulated in nitrogen‐treated plants at 48 h (log₂FC = 2.7, FDR = 0.03) (Figure 4c and Table 6).

To explore whether nitrogen resupply could impact dhurrin turnover, we also analyzed the expression of genes involved in dhurrin recycling. While the glutathione‐S‐transferase (GST) responsible for catalyzing the conversion of dhurrin to the GS‐pOHPACN conjugate remains unidentified, there is a GST encoded by Sobic.001G012500 (GST1) that is co‐located with the dhurrin biosynthesis genes in the chromosome 1 gene cluster, which is a candidate for catalyzing the first step in dhurrin recycling (Bjarnholt et al. 2018). A GST1 transcript, Sobic.001G012500.2, was upregulated in leaf tissue of nitrogen‐treated plants at four time points: 12 and 36–60 h, with around 20‐ to 25‐fold greater than in controls from 36 to 60 h (log₂FC = 3.9–4.6, FDR < 0.001). Two lambda‐class glutathione‐S‐transferases (GSTLs) (Sobic.002G421200 [GSTL1] and Sobic.009G033200 [GSTL2]) and two nitrilases (NIT) (Sobic.004G225100 [SbNIT4B2] and Sobic.004G225200 [SbNIT4A]) are likely to be involved in later stages of the pathway (Jenrich et al. 2007; Nielsen et al. 2016; Bjarnholt et al. 2018). These recycling‐related genes are expressed in leaf tissue of both nitrogen‐limited control and nitrogen‐treated plants (Table 4), but there was no clear change in their expression in response to nitrogen resupply (Table 4). Similarly, expression patterns in stem (Table 5) and root tips (Table 6) did not show a clear response to nitrogen application.

There are differences in the expression of genes involved in bioactivation of dhurrin. In leaf tissue, SbHNL was significantly upregulated from 36 h onwards (log₂FC = 7.9, FDR < 0.01), but no changes in expression of dhurrinase genes (SbDHR1 and 2 and SbDHR‐like3 and 4) responsible for dhurrin hydrolysis were observed (Table 4).

2.9. GRN Analysis Identified Putative Nitrogen‐Responsive Regulators of Dhurrin Biosynthesis

As leaf tissue was the only tissue where the dhurrin biosynthesis genes were DE, a series of GRNs was constructed, including time points 24, 36, 48, and 60 h, to investigate the nitrogen‐responsive networks controlling dhurrin production in leaf tissue (Figures S13–S16). These networks included many orthologues of known nitrogen‐responsive TFs, such as AtHRS1, known in rice as Nitrate Inducible GARP‐type Transcriptional Repressor1 (OsNIGT1) (Figure S14, 36 h, and Figure S16, 60 h), AtLBD37 (Figures 13–15, 24–48 h), and TGACG motif‐binding factor 4 (AtTGA4) (Figure S16, 60 h). The functions of these TFs have been characterized predominantly in the root tissue of arabidopsis and rice but likely have similar roles in coordinating nitrogen responses in leaves and across different species. Many genes with predicted functions in nitrogen metabolism were present in the leaf networks, with some predicted to occur in the same regulatory modules as the dhurrin‐related genes. For example, at 24 h in leaf tissue, SbCYP71E1 and SbMATE2 are both predicted to be upregulated by Sobic.007G210700, an orthologue of Ethylene‐Insensitive‐like3 (AtEIL3) (Figure S13, 24 h). SbEIL3 was also predicted to activate expression of orthologues for AtNRT1.13 and AtAVT6C, involved in nitrate sensing and amino acid transport, respectively (Figure S13, 24 h). Similarly, Sobic.003G136600, an orthologue of AtMYB61, was a predicted regulator of SbCYP79A1 and SbCYP71E1, as well as SbNADH‐GOGAT, SbGS2, and SbUPM1 which are key enzymes and cofactors for inorganic nitrogen assimilation (Figure S13, 24 h).

Subnetworks including the direct regulators of genes related to dhurrin metabolism were extracted from the GRNs at each of the four selected time points (Figures S17–S20). SbCYP79A1 catalyzes the rate‐limiting step of the biosynthesis pathway, and three candidate regulators of SbCYP79A1 expression were identified in the leaf GRNs at different time points, all of which are predicted to be negative regulators of SbCYP79A1 expression, as their expression is negatively correlated. An orthologue of the MYB‐family TF AtMYB61, Sobic.003G136600, was predicted to regulate SbCYP71E1 expression at 24 h (Figure S17) and also SbUGT85B1 at 36 h (Figure S18). Two members of the bZIP TF family, Sobic.003G32700 and Sobic.003G050500, were predicted to regulate SbCYP79A1 expression at 36 h (Figure S18) and at 48–60 h, respectively (Figures S19 and S20). Sobic.003G32700 is an orthologue of AtbZIP23 and is also predicted to regulate SbCYP71E1 and SbUGT85B1 at 36 h (Figure S18), while Sobic.003G050500 regulates all three biosynthesis genes at 48 and 60 h (Figures S19 and S20).

Other predicted regulators of SbCYP71E1 and/or SbUGT85B1 expression include Sobic.001G078900 (HRS1, G2‐like family) (Figures S18 and S20, 36 and 60 h), Sobic.005G177800 (POPEYE (PYE), bHLH family) (Figure S20, 60 h), Sobic.007G210700 (EIL3) (Figure S17, 24 h), and Sobic.002G422000 (MYB‐related protein 1 (MYR1), G2‐like) (Figure S18, 48 h). Several direct regulators of the dhurrin biosynthesis genes present in the leaf networks are predicted to interact with more than one dhurrin metabolic process (Figure 5). For example, bZIP family TF Sobic.003G050500 is predicted to regulate genes involved in biosynthesis, bioactivation, detoxification, transport, and recycling. Interestingly, Sobic.003G050500 expression is negatively correlated with dhurrin biosynthesis genes but positively correlated with the expression of SbNIT4B2, which functions in dhurrin turnover and recycling.

FIGURE 5.

Summary of direct interactions between transcription factors and target dhurrin metabolism genes predicted in gene regulatory networks (GRNs) at 24, 36, 48 and 60 h after nitrogen treatment. Gene IDs for transcription factors listed along the top of the plot, with the transcription factor family and gene symbol for the Arabidopsis thaliana orthologue given in brackets where known. Gene symbols for dhurrin metabolism target genes and the GRN time point from which they originate are listed along the left of the chart. Interactions were filtered for mutual rank (MR) > 0.7. Cell color indicated the direction and strength of the correlation (Pearson's correlation coefficient, PCC) of the expression (TPM) between a transcription factor and its target node.

At higher levels of the network are many NAC‐family TFs, bHLH‐family members, and several regulators involved in hormone responses, such as abscisic acid (ABA) (ANAC029 [Sobic.005G018500]; MYB15 [Sobic.006G148000]; ethylene response factor 48 [ERF48] [Sobic.009G101400]), auxin (auxin response factor 8 [ARF8] [Sobic.004G051900] MYR1 [Sobic.002G422000]), and cytokinin (arabidopsis response regulator 12 [ARR12] [Sobic.010G064700]; HECATE 2 [HEC2] [Sobic.007G155300]) (Figures S17–S20). Some of these Arabidopsis orthologues for putative regulators are also associated with senescence and developmental maturation, for example, AtERF48, AtANAC029, and AtMYB59 (Figures S17–S20). One or more members of the ERF, MYB, and NAC families of TFs, frequently associated with senescence, are present in all subnetworks constructed (Figures S17–S20).

3. Discussion

Cyanogenic glucosides have important roles in plant defense. They may also function as nitrogen storage compounds via a recycling pathway (Gleadow and Møller 2014; Bjarnholt et al. 2018; Gleadow et al. 2021; Kliebenstein 2023), but experimental evidence is limited. It is well‐established that nitrogen influences the amount of dhurrin in sorghum ( S. bicolor ) (Boyd et al. 1938; Blomstedt et al. 2018; Myrans et al. 2021). Nitrogen also induces expression of the dhurrin biosynthesis genes in plants after the early seedling stage (Busk and Møller 2002). Here, we investigated the regulatory responses driving expression of the dhurrin biosynthesis pathway genes in response to nitrogen resupply. We found that dhurrin biosynthesis was upregulated in response to nitrogen resupply in leaves but not in stems or root tips. By constructing GRNs, we identified putative regulators of the dhurrin pathway in leaf tissue, including those with functions in nitrogen signaling pathways. Finally, we relate the findings to the importance of increased nitrogen use efficiency coupled with the need for forage that remains safe under different environmental conditions.

3.1. Global Impact of Nitrogen Limitation and Response to Nitrogen Treatment

The plants in the present experiment were dosed with a high nitrogen‐containing solution at the 6‐leaf stage, after a period of growth under nitrogen limiting conditions and harvested at intervals. Plants that were harvested before the addition of extra nitrogen, and the controls, had lower chlorophyll concentrations and lower biomass accumulation than generally observed in sorghum grown under optimal conditions (Figures 1d, S2, and S3). These characteristics are typical of nitrogen deficiency (Masclaux‐Daubresse et al. 2010; Xu et al. 2012) and are consistent with other work where sorghum was grown under low nitrogen (Zhao et al. 2005; Gelli et al. 2014). Nitrogen limitation reduces the capacity for amino acid, protein, and nitrogenous metabolite production and the capacity for light harvesting and carbon assimilation (Zhao et al. 2005). The disruption of central carbon and nitrogen metabolism is a key factor in reducing dry matter accumulation and grain yield in sorghum under nitrogen limiting conditions (Gelli et al. 2014; Worland et al. 2017; Zhu et al. 2020; Choi et al. 2024). Plants treated with 12 mM of nitrogen solution showed no increase in growth‐related parameters within the 60‐h experimental window, with similar height, biomass, and leaf stage recorded regardless of treatment across all timepoints (Figure S3). This result reflects the time required for nitrogen uptake, assimilation, and translation into a measurable growth increase and is not unexpected within the short experimental window. While sorghum is known for having higher nitrogen use efficiency than many other crops (Bollam et al. 2021), optimal nitrogen nutrition remains critical to support normal growth and development. Thus, improving nitrogen use efficiency is a key goal of many sorghum breeding programs (Massel et al. 2016; Ostmeyer et al. 2022).

Many changes were detected in the transcriptional profiles of root, stem, and leaf tissues upon nitrogen treatment (Figure 2), indicating a response to the sudden increase in nitrogen availability. Changes in the transcriptome were largely organ‐specific, with distinct patterns of gene expression observed for leaf, stem, and root tips—especially regarding the timing of the transcriptional response. For example, the earliest changes in gene expression were detected in roots just 2 h after treatment. Many of these genes were directly related to nitrogen metabolism, with increased expression of genes in both the nitrate and ammonium assimilation pathways (Table 1), and several nitrogen transporters, including an orthologue for AtNRT1.1, with a dual function in nitrate transport and sensing (Ho et al. 2009; Liu et al. 2017), and AtCLC‐A (Figure 3b), with a role in vacuolar transport of nitrate (Monachello et al. 2009).

Nitrate concentration was significantly higher in roots from 6 h and sheath from 12 h after nitrogen application (Figure 1b,c), consistent with increased expression of the nitrogen transport and assimilation pathways, and indicative of a progressive translocation of nitrate from roots to shoots. There was no change in the nitrate concentration in leaf tissue in response to nitrogen treatment. The leaf is a key site for nitrate assimilation, which is dependent on reducing power generated by photosynthesis. The lack of nitrate accumulation in leaf tissue is consistent with the widely recognized observation that nitrate is not stored in the leaves, unlike the stem, but is directly assimilated and channeled into primary metabolism. This is supported by the observed upregulation of several genes involved in primary nitrogen assimilation in leaves from 2 to 6 h after nitrogen application (Table 1). GO term enrichment in leaf tissue also pointed to the activation of downstream metabolic processes, including photosynthesis and amino acid biosynthesis (Figure S6), and is consistent with the increase in chlorophyll at later time points.

Analysis of GRNs in the present study identified candidate transcriptional regulators with similar functions in A. thaliana and the model crops maize ( Zea mays ) and rice ( Oryza sativa ). Transcriptional nitrogen responses have mainly been investigated in roots, and several TF families with roles in activation and suppression of nitrogen responses have been characterized (Rubin et al. 2009; Krapp 2015; Gaudinier et al. 2018). Key components of the nitrogen signaling pathway include NRT1.1, NLP6/7, TGA4 (Marchive et al. 2013; Alvarez et al. 2014; Liu et al. 2017), of which several orthologues were identified in the GRNs created for 12 h root and 24–60 h of leaf data. While no nitrogen‐responsive regulators have been functionally characterized for sorghum, there is support for conservation of components of the nitrogen signaling pathway. For example, Du et al. (2020) studied early responses to nitrogen in maize and sorghum and identified a conserved set of genes responsive to nitrogen supply, and many of the identified transcripts identified in roots in the present study were also orthologous to nitrogen‐related signaling components identified in Arabidopsis. These results validate the approach detailed here to identify relevant candidates for regulation of genes and pathways that are responsive to changes in plant nitrogen availability, specifically in dhurrin biosynthesis and recycling.

3.2. Dhurrin Accumulation Is Highest in Developmentally Young Tissue

In this study, the stem had approximately 10 times higher HCNp than leaf, sheath, or root tissues (Figure 1e–h). The relatively high HCNp in stem tissue compared to the leaf and sheath may be related to differences in organ function or the stage of tissue development (Miller et al. 2014). At the time of harvest, the plants in the present study were in the juvenile stage of growth (5–6 leaf stage) and samples for leaf and sheath were collected from fully expanded (mature) leaves. However, stem tissue in plants at this growth stage is enriched in meristematic tissue (Yu et al. 2022). For example, while internode growth is suppressed during the juvenile phase, the lower stem nodal tissues are producing crown roots; nodes closer to the shoot apical meristem are producing new leaves, and a new node is produced by the shoot apical meristem every 3–4 days. Previous studies have shown that dhurrin pathway genes are highly expressed in the stem apical growing zone (Gleadow et al. 2021), consistent with high HCNp in stems in the current study.

HCNp concentrations in leaf, sheath, and roots under nitrogen limitation were much lower than those in stem. These findings are consistent with previous studies, where long‐term growth under low nitrogen conditions is correlated with reduced dhurrin concentrations (Kriedman 1964; Myrans et al. 2021). In the present study, leaf and root HCNp for sorghum plants at the 6‐leaf stage were comparable to 8‐ to 10‐leaf stage plants described in Blomstedt et al. (2018), which were developmentally older but grown under a higher 2‐mM nitrogen regime, suggesting that dhurrin declined more rapidly under nitrogen stress. However, the fact that HCNp was measurable even under long‐term nitrogen stress supports the idea of a basal dhurrin level in vegetative tissues, which has been suggested previously for sorghum (Blomstedt et al. 2018) and also for cyanogenic Eucalyptus cladocalyx (Simon et al. 2010).

3.3. Response of Dhurrin to Nitrogen Application Is Dependent on Tissue Type

Nitrogen resupply was associated with increased expression of the dhurrin biosynthesis pathway in leaf tissue, with SbCYP79A1 and SbCYP71E1 upregulated from 24 h following the nitrogen treatment and SbUGT85B1 upregulated from 36 h after nitrogen treatment (Figure 4a). Nitrogen‐mediated upregulation of the dhurrin biosynthesis pathway is consistent with earlier work (Busk and Møller 2002), as is the timing of the upregulation response within 24 h of an exogenous treatment (Sohail et al. 2023). In the leaf tissue of nitrogen‐treated plants sampled prior to 24 h and in the nitrogen‐limited control plants at all time points, expression of the biosynthesis pathway genes was not detected. Lower expression of the dhurrin biosynthesis pathway is commonly observed in older plants, as the pathway is switched off as plants mature (Blomstedt et al. 2018; Rosati et al. 2024). The HCNp detected in leaf tissue in the absence of biosynthesis pathway gene expression likely reflects both the stability of the dhurrin compound and the pathway enzymes, which may be supported by their organization within a metabolon (Laursen et al. 2016) or their storage in a specialized natural solvent (Knudsen et al. 2020).

The dhurrin biosynthesis genes were expressed in stems and root tips, although nitrogen resupply had no impact on their expression patterns, which was consistent with the results for HCNp. HCNp in stems was already high, and possibly at maximal capacity, indicating that nitrogen treatment was unable to induce dhurrin production further. Similar reports have been made for sorghum seedlings, where nitrate does not induce expression of the biosynthesis genes in plants less than 8‐day post germination (Busk and Møller 2002), and dhurrin concentrations in younger plants are less susceptible to other environmental conditions, such as osmotic stress (O'Donnell et al. 2013). A similar effect may explain the lack of increase in dhurrin biosynthesis gene expression in root tips, which also contain nondifferentiated meristematic tissue. Additionally, dhurrin concentration in roots is generally less susceptible to environmentally induced increases, such as water limitation and other nutrient variations (O'Donnell et al. 2013; Miller et al. 2014) but can be perturbed under extended growth under high nitrogen regimes (Blomstedt et al. 2018; Myrans et al. 2021). It is also possible that the strength or duration of the nitrogen treatment was insufficient to induce expression of the dhurrin biosynthesis pathway in these tissues.

Taken together, these results suggest that dhurrin biosynthesis is regulated differently in leaves compared to stems and root tips and more strongly regulated by environmental signals in leaves than in other tissues, as proposed by Gleadow et al. (2021). These differences in expression patterns may result from tissue‐specific regulatory mechanisms or could be a result of different metabolic demands of developmentally younger organs, for example, leaves compared to stems.

3.4. Dhurrin and Expression Profiles Varied With Time of Day

Dhurrin also varied with time of day: HCNp was higher in nitrogen‐treated plants at the evening time points (36 and 60 h after nitrogen application) in both leaf and sheath. In contrast, the expression profiles followed an opposing diel pattern: transcripts of SbCYP79A1, SbCYP71E1, and SbUGT85B1 peaked at morning time points (24 and 48 h) and then dipped in the evening (36 and 60 h) (Figure 4a). Notably, the higher HCNp in leaf tissues in the evening was driven, in part by a relative reduction in HCNp in the controls rather than increased concentration in response to nitrogen treatment alone.

While there was no clear trend in downregulation of genes related to dhurrin recycling in response to nitrogen treatment (SbNIT4B2, SbGSTL1, SbGSTL2), the pathway appeared to be stably expressed, suggesting constant recycling of dhurrin. Coupled with diurnal regulation of the biosynthesis pathway, dhurrin turnover sustained overnight could explain the decrease in HCNp at the 48‐h time point, as well as the steady decline in HNCp observed in control plant leaf tissues at the later time points. Adewusi (1990) also found that dhurrin concentrations were a result of the balance between biosynthesis and turnover; however, their results indicate increased biosynthesis at night. As dhurrin biosynthesis is strongly influenced by plant developmental stage (Miller et al. 2014), this difference could be related to the age of the plants used, as Adewusi (1990) used seedlings less than 1 week old. Therefore, it is possible that regulation of biosynthesis and turnover may be different in these very young plants.

Diurnal expression of the dhurrin biosynthesis genes has been previously described for leaf tissue in sorghum (Gleadow et al. 2021). Cyanogenic glucosides produced in cassava ( Manihot esculenta ) also show diurnal expression patterns (Schmidt et al. 2018). Diurnal regulation could be a mechanism to coordinate dhurrin biosynthesis with resource availability, as during the day photosynthesis can be coupled with anabolic processes, including nitrogen assimilation and metabolite production (Pérez‐Llorca and Müller 2024; Sathee et al. 2025). The upstream region of SbCYP79A1 contains regulatory elements associated with light signaling (Fromhold 2015), and further investigation into the circadian regulation of dhurrin biosynthesis could help to understand how this is mediated.

3.5. Regulatory Networks Identify Candidates for Control of Dhurrin Biosynthesis and Suggest Suppression of Dhurrin Biosynthesis Under Nitrogen Limitation

Knowledge of the regulatory mechanisms controlling the biosynthesis of dhurrin lags behind the understanding of its biochemistry. In this study, GRNs were constructed for sorghum leaf tissue to identify candidate TFs regulating dhurrin biosynthesis in response to nitrogen resupply. Recently, Rosati et al. (2024) identified SbGATA22 as a regulator of dhurrin biosynthesis in sorghum leaves, mediating repression of SbCYP79A1 with plant maturity. Though some GATA TFs are involved in mediating nitrogen responses (Chiang et al. 2012), SbGATA22 was not identified in the GRN in this study as there was no significant difference in expression between control and nitrogen treated tissue. This result was not unexpected, as SbGATA22 is associated with repression of the pathway during the transition from seedling to juvenile, and may not be a key regulator of dhurrin at later developmental stages (Rosati et al. 2024). To our knowledge, there are no published papers implicating other TFs in regulating dhurrin biosynthesis.

Positive transcriptional regulators predicted to regulate the dhurrin biosynthesis pathway were from diverse families: G2‐like (e.g., SbHRS1, SbMYR1), ethylene‐insensitive like (e.g., SbEIL3), and MYB‐related (e.g., SbTRFL3). None were predicted to induce expression of SbCYP79A1, although as the genes for the biosynthesis pathway are clustered, it is possible that these transcriptional activators play a role in co‐regulating expression of the pathway genes; however, these interactions were not detected in this study. The G2‐like family TF AtHRS1/OsNIGT1 has a conserved role in regulating the response to nitrogen in Arabidopsis and rice ( O. sativa ) (Sawaki et al. 2013; Medici et al. 2015) and likely coordinates nitrogen responses in sorghum as well (Du et al. 2020). Regulation of SbCYP71E1 expression by SbHRS1 (Sobic.001G078900) predicted in the 36‐ and 60‐h GRNs (Figure 5) supports a mechanism for coordinating dhurrin biosynthesis with nitrogen supply. Co‐opting nitrogen signaling network components to the regulation of dhurrin biosynthesis would enable synchronization of dhurrin production with nitrogen availability. Further work to characterize the putative regulatory components will provide greater insight into nitrogen‐mediated upregulation of the dhurrin biosynthesis pathway.

Sobic.003G136600, an orthologue for AtMYB61 was predicted to be a negative regulator of dhurrin biosynthesis in the GRNs, since downregulation of this transcript in nitrogen‐treated plants was correlated with upregulation of the biosynthesis genes at 24‐ and 48‐h time points. Upstream regulators of Sobic.003G136600 at the 24‐h time point include two orthologues of AtLBD37, a member of the LOB‐Domain family: Sobic.001G328500 and Sobic.002G359700. LBD37/38/39 are known as suppressors of nitrogen‐deficiency responses in Arabidopsis and rice (Rubin et al. 2009; Zhu et al. 2022). In sorghum, LBD37 orthologues may repress Sobic.003G136600 expression in response to nitrogen signaling, thereby enabling upregulation of the dhurrin biosynthesis pathway under nitrogen‐sufficient conditions.

Senescence, the final stage in leaf development, is characterized by the breakdown of cellular components and recovery of stored nutrients to support growth of sink tissues (e.g., newly developing leaves, grain fill). Several senescence‐associated TFs were identified across the leaf GRNs and the subnetworks for the dhurrin biosynthesis pathway genes analyzed. These included orthologues for members of the NAC (AtANAC047, AtANAC087, and AtANAC029), WRKY (AtWRKY22), and ERF (AtERF48) families (Balazadeh et al. 2008; Zhou et al. 2011; Shi et al. 2023; Van Durme et al. 2023). MYB‐family TFs were also present in the subnetworks, including orthologues for AtMYB15 and AtMYB59, both known to interact with ABA and salicylic acid (SA) signaling pathways to promote senescence (Ding et al. 2009; Wang et al. 2021; He et al. 2023). Early senescence can be induced under nitrogen limitation (Havé et al. 2016; Chen et al. 2020), but nitrogen resupply can reverse nitrogen‐limitation‐induced senescence programs (Balazadeh et al. 2014). In the GRNs, the senescence‐associated genes were mostly downregulated, suggesting that under prolonged nitrogen limitation, the dhurrin biosynthesis pathway is strongly suppressed in leaves, but reversal of the senescence program enables derepression of the biosynthesis pathway genes.

The delayed senescence phenotype known as stay‐green is associated with higher yields under postanthesis drought stress and improved nitrogen use (Borrell and Hammer 2000; Xu et al. 2012; Borrell et al. 2014; Varoquaux et al. 2019). Stay‐green in sorghum is also positively correlated with dhurrin concentration (Hayes et al. 2023), and dhurrin content has been used as a marker for stay‐green lines (Burke et al. 2013; Hayes et al. 2016). Gleadow et al. (2021) proposed that dhurrin in the flag leaf may be broken down to release nitrogen during grain filling, delaying the need to remobilize Rubisco. However, dhurrin does not appear to contribute directly to the stay‐green phenotype, as the stay‐green phenotype is still apparent in a dhurrin‐free mutant line (Gruss et al. 2023). Nevertheless, it is possible that the high dhurrin concentrations in stay‐green lines may be due to prolonged dhurrin biosynthesis associated with delayed senescence programs.

4. Conclusion

Here, we identified candidate regulators of dhurrin biosynthesis in response to nitrogen application and revealed putative mechanisms for nitrogen‐mediated control of dhurrin production in sorghum. The influence of diel regulation on the expression of dhurrin biosynthesis pathways and dhurrin production was highlighted, with possible implications for the role of dhurrin as a nitrogen storage compound. High and variable dhurrin production is dangerous where sorghum is used for forage, as high concentrations of dhurrin can lead to HCN poisoning of livestock. These findings could inform breeding programs to develop forage varieties with more predictable dhurrin concentrations. Improving understanding of the regulatory networks controlling both dhurrin biosynthesis and nitrogen metabolism in sorghum could provide targets for improving dhurrin accumulation patterns and nitrogen use efficiency in sorghum. Further characterization of putative regulatory components identified in the regulatory networks constructed during this analysis could reveal how dhurrin biosynthesis is controlled and how it fits within regulation of plant metabolism more broadly.

5. Experimental Procedures

5.1. Plant Growth Conditions and Experimental Design

Seeds of S. bicolor cv. BTx623 were germinated in water at room temperature and planted in sand (Bastion River Sand, Dingo Cement) and allowed to establish in the greenhouses at Monash University, Clayton, Australia, in January 2022. At the 1‐leaf stage, seedlings were transferred to 10‐cm pots containing sand and maintained at 18°C–28°C under natural photoperiod (January–March), with water delivered twice a day with a drip irrigation system.

For growth under limiting nitrogen conditions, plants received 200 mL of a modified 0.5 mM of Nitrogen Long Ashton's Nutrient Solution (N LANS) three times per week containing the following: 2 mM of K₂SO₄, 1.5 mM of Mg SO₄, 2 mM of K₂HPO₄, FeNaEDTA 0.04 g L⁻¹, 3 mM of CaCl₂, 3.9375 mM of Na₂SO₄, 0.0625 mM (NH₄)₂SO₄, 0.375 mM NaNO₃, and micronutrients at the following concentrations: H₃BO₃ 2.86 mg L⁻¹, MnCl₂.4H₂O 1.81 mg L⁻¹, ZnSO₄.7H₂O 0.22 mg L⁻¹, CuSO₄.5H₂O 0.08 mg L⁻¹, Na₂MoO₄.2H₂O 0.025 mg L⁻¹. The concentrations of NaNO₃ and (NH₄)₂SO₄ were reduced to contain a total of 0.5 mM N in a ratio of 6:1 (mol:mol) NO₃ ⁻ to NH₄ ⁺ and salts balanced by adjusting Na₂SO₄ according to Miller et al. (2014).

Plants were maintained at 0.5 mM N of LANS until they reached the 6–7 leaf stage, when treatments were imposed. Half the plants were watered with 200 mL of LANS modified to contain 12 mM N (nitrogen‐treated), while the remainder received a LANS without NO₃ ⁻ or NH₄ ⁺ (0 mM N, control). Modified 12 mM N of LANS contained 2.5 mM of Na₂SO₄, 1.5 mM of (NH₄)₂SO₄, and 9 mM of NaNO₃, while the 0 mM N of control solution contained 4 mM of Na₂SO₄ but no (NH₄)₂SO₄ or NaNO₃.

5.2. Harvest Protocol

Plants were destructively harvested at intervals with plants assigned to be sampled for chemical (n = 4) or transcriptomic analysis (n = 5) at each time point. Tissue for the baseline harvest (0 h, before treatments were imposed) was collected on March 7, 2022. The following day at 8 a.m., the treatment solution (12 mM N) was delivered to half the plants, and the control solution (0 mM N) was applied to the remaining plants. Individual plants were randomly selected from control (all time points) and surge treatments (from 2h onwards) to be harvested at each time point (Figure S1a).

Plants designated for biomass determination and chemical analysis were divided into leaf blade (leaf hereafter), leaf sheath (sheath, hereafter), true stem (stem hereafter), and roots and dried in paper bags in an oven at 60°C for 72 h and weighed to determine dry mass (Figure S1b). Plant material was ground to a fine powder using a bead mill (TissueLyser II, Qiagen, Hilden, Germany) for chemical analysis (see below).

For transcriptomic analysis, samples were taken from the midsection of the sixth leaf, the stem, and root tips (Figure S1b) and immediately frozen in liquid nitrogen and stored at −80°C. Remaining tissue (leaves, sheaths, and roots) was oven‐dried at 60°C, ground, and analyzed for nitrate concentration and HCNp.

5.3. Morphological Measurements

Prior to harvesting at each time point, leaf number (i.e., the total number of fully emerged leaves on a plant, with a leaf judged as fully emerged once the ligule is visible) and height (mm) (from the base of the pot to the ligule of the last fully emerged leaf) were determined. The weights of the oven‐dried tissues were recorded to determine dry mass for leaf, sheath, and roots.

5.4. Chlorophyll Concentration

Prior to destructive harvest, chlorophyll concentration (μmol m⁻²) was estimated using a SPAD meter (MC‐100, Apogee Instruments. UT, USA) on the 6th leaf of each plant, corresponding to the tissue sampled for transcriptomic analysis. Custom measurement mode was selected, with “sorghum” chosen as the species to enable accurate conversion of readings to chlorophyll concentration (μmol m⁻²). Each measurement is an average of three readings taken at points along the mid‐section of the blade, avoiding the midrib and excluding the tip and leaf base.

5.5. Chemical Analyses

Nitrate (NO₃ ⁻) concentration of each tissue was determined using a modified version of the colorimetric method described by Cataldo et al. (1975). Finely ground oven‐dried plant material (15–16 mg) of leaf, sheath, and root tissues was extracted with 500 μL of RO water (in a bead mill, 2 × 2 min, 28 Hz, TissueLyser II, Qiagen) and incubated at 45°C for 1 h with regular mixing by inversion. After incubation, samples were centrifuged for 15 min (approximately 16,000 × g) and the supernatant collected. Briefly, 5 μL of sample was added to a 96‐well plate and incubated with 20 μL of 5% SA in concentrated H₂SO₄ for 20 min at room temperature. After incubation, 300 μL of 3 M NaOH was added to each well to neutralize the sample. Absorbance was read at 410 nm on a plate reader (FLUOstar Galaxy, BMG LABTECH, Ortenberg, Germany). Samples were analyzed in triplicate, and mean absorbance values were used to determine NO₃ ⁻ as mg g⁻¹ dry weight tissue through comparison with a KNO₃ ⁻ standard curve.

HCNp is the total amount of cyanide evolved from cyanogenic glucosides present in a given tissue sample after addition of exogenous hydrolytic enzyme, expressed as the concentration of HCN mg g⁻¹ plant tissue. The HCNp of dried and ground leaf, sheath, stem and root tissues was determined following Gleadow et al. (2012). In brief, ground plant material (9.5–10 mg) was incubated with 300 μL of 0.2% w/v β‐glucosidase enzyme (G‐0395, Sigma‐Aldrich, MO, USA) in 0.1 M of citrate buffer (pH 5.6, trisodium citrate, Sigma‐Aldrich). A 0.2 mL of microcentrifuge tube containing 200 μL of 1 M NaOH was inserted and vials tightly sealed. Tissue disruption was achieved by freezing the vials at −20°C and rethawing for 1–2 h at room temperature with gentle agitation (60 rpm on orbital shaker). Vials were then incubated at 37°C overnight (16–20 h) to allow for the complete degradation of cyanogenic glucosides and collection of volatile HCN in the NaOH vial. The NaOH/NaCN vial was removed and stored at 4°C prior to analysis. Determination of HCN in samples was done with 5 μL of each NaOH/NaCN sample diluted with 45 μL of distilled water on a 96‐well plate. For quantification using the colorimetric assay, 50 μL of 0.5 M of acetic acid, 125 μL of succinimide reagent, and 50 μL of barbituric acid in pyridine reagent were added to each well. Following 15‐min incubation at room temperature, absorbance of each sample well was measured at 595 nm on a plate reader (FLUOstar Galaxy, BMG LABTECH). Samples were analyzed in triplicate, with mean absorbance used to determine HCNp as mg g⁻¹ dry weight through comparison with a NaCN standard curve.

5.6. RNA Sequencing and Transcriptomic Analysis

Tissue for RNA extraction was ground to a fine powder with a bead mill (FastPrep‐24, MP Biomedicals, CA, USA) using the cryogenic attachment cooled with dry ice. RNA was extracted using the Spectrum Plant Total RNA Extraction Kit (Sigma‐Aldrich) and treated with the on‐column DNase digest (DNASE70, Sigma‐Aldrich) to remove contaminating genomic DNA. RNA 260/280 and 260/230 ratios were checked with a NanoDrop 1000 spectrophotometer (ThermoFisher Scientific, MA, USA), quantified with a fluorometer (Qubit 4, Invitrogen, MA, USA) with RNA Broad Range sensitivity kit (Q10210, Invitrogen), and integrity checked on a 2% agarose gel. Additionally, 10% of samples were checked on a BioAnalyzer 2100 (Agilent, CA, USA) to ensure RIN > 6.0 in line with minimum quality standards at the Joint Genome Institute (JGI) where samples were sequenced to a depth of 30–50 m reads per replicate using Illumina short read sequencing. Transcriptome assembly was performed as described previously (Yu et al. 2022). Reads were aligned to the S. bicolor reference genome v3.1 (McCormick et al. 2018) using the HISAT2 aligner. Transcriptome assembly and transcripts per million (TPM) normalization were performed with StringTie v1 (Pertea et al. 2015). The script prepDE.py https://github.com/gpertea/stringtie/blob/master/prepDE.py and https://ccb.jhu.edu/software/stringtie/index.shtml?t=manual, was used to convert nucleotide coverage data from StringTie into read counts that are used by differential expression statistical packages. Pair‐end read length was 151 bp per read. Functional annotations of the transcripts were obtained from the S. bicolor V3.1 (McCormick et al. 2018).

Prior to differential gene expression analysis, exploratory data analysis was conducted to check sample quality. Samples were subject to library size normalization and then filtered to remove lowly expressed transcripts (CPM > 0.5 in two or more samples) and further TMM normalized using the edgeR package (Chen et al. 2025). Boxplots and PCA plots were generated to check library distribution and replicate consistency, respectively (see Figure S4a–d).

Four root samples did not meet quality criteria at JGI and were not sequenced: two replicates of the 36‐h nitrogen treatment, one 6‐h root control replicate, and one 60‐h root nitrogen replicate. In addition, PCA analysis indicated one 24‐h root nitrogen sample clustered with the replicates for 24‐h root control, and this sample was excluded. The 36‐h root condition was excluded from the analysis due to insufficient replicates. The 6‐h control, 24‐h nitrogen, and 60‐h nitrogen conditions for root tissue were retained in the analysis with two replicate samples for their respective treatment and time point combination.

Differential gene expression analysis was performed on transcript‐level gene counts, and DE transcripts were identified using the edgeR package. As no comparisons between tissues were planned for downstream analysis, each tissue was analyzed separately to reduce variation in expression driven by tissue type rather than treatment and time point. Contrasts were made between treatment and control samples at each time point (2–60 h). Results were filtered to retain genes with fold change (FC) ≥ |5|, false discovery rate (FDR) < 0.05, and TPM ≥ 5. Where FC is reported within the text, these are the values calculated within the edgeR analysis except in cases where TPM for either treatment is less than 1. In these cases, for the treatment near to zero, a value of one was added to the TPM to prevent reporting misleading FC values: FC = TPM N/ (TPM C + 1). Clustering of DE transcripts in each organ type was done using hierarchical clustering based on Euclidean distances between transcripts and samples using the complete linkage method. Clustering was performed on z‐score transformed means of log₂₋transformed TPM expression values.

GO enrichment analysis was performed using the GO Term Enrichment tool at Plant Transcriptional Regulatory Map (PlantRegMap, https://plantregmap.gao‐lab.org/index.php, Tian et al. 2020). Settings used were pecies: S. bicolor , aspect: biological process (P), p < 0.05. GO enrichment analysis was performed for unique transcripts DE in at least one tissue at any time point (2–60 h) and results filtered at FDR < 0.01. GO enrichment was also performed on the upregulated and downregulated DE genes within each tissue and time point combination (i.e., leaf, 24‐h, upregulated), and significantly enriched GO terms were further filtered at FDR < 0.1. Obsolete GO terms present in the PlantRegMap output were either updated to current terms or excluded from the analysis. Jaccard similarity (JS) analysis was performed to identify groups of GO terms with significant overlap of genes (> 40%) within each tissue and time point combination. GO terms with significant overlap were grouped into a single category and manually annotated with the most representative GO term for that group (see Tables S4–S6).

GRN analysis was performed as previously described (Yu et al. 2022). Each GRN was constructed using a list of DE transcripts for a tissue and time point combination (i.e., leaf, 24‐h), including both upregulated and downregulated transcripts. Expression values (TPM) for the list of DE transcripts were retrieved for all samples in the dataset regardless of time point and treatment and were included in the calculation of a correlation matrix (Pearson's correlation coefficient, PCC). Where multiple transcripts of the same gene were present, TPM was summed across all transcripts so that only one set of expression values for each unique gene remained in the matrix. Mutual rank (MR) for each gene pair was also calculated. To connect TFs in the network with putative target genes, enriched transcription factor binding sites (TFBS) within putative promoters were retrieved for each gene in the dataset (> 1.1‐fold enrichment). The TFBS information determined by Yu et al. (2022) was used for filtering the TFs in the present analysis. Briefly, putative promoters were defined as the 1‐kb region upstream of the transcription start site (TSS). TSS information was obtained from the Morokoshi sorghum database (Makita et al. 2015). The sequences of the putative promoter regions were scanned using the Plant Promoter Analysis Tool available on PlantPAN 4.0 (Chow et al. 2019), and TFBS were calculated using position weight matrices for all genes in the sorghum genome. Genes without enriched TFBS for TFs present in the dataset were excluded. Filtering during GRN construction mainly removed genes not classed as TFs from the datasets; however, in each of the Leaf 24‐ and Leaf 48‐h GRNs, one TF was excluded from input sets of 13 and 29 TFs, respectively (Tables S12 and S14). Remaining genes in the network were visualized using Cytoscape software v.3.10.1 (Shannon et al. 2003) and trimmed further using PCC and MR values on edges (indicated in plots). This filtering removed three TFs from the Root 12‐h GRN, from an original set of 46 TFs (Table S16). For the leaf GRNs, filtering of edges was done to maximize the number of dhurrin metabolism genes in the network as the goal of the analysis was to identify regulators of the dhurrin biosynthesis pathway. As such, no PCC filtering was done on these networks to avoid excluding genes in the biosynthesis pathway. Networks were analyzed with the analyzer tool within cytoscape and edge counts generated for each node (analyzed as a directed graph). Subnetworks containing the dhurrin biosynthesis pathway genes were constructed by selecting the biosynthesis genes present in the network and their direct regulators, plus two additional levels of regulators. The final subnetworks therefore included in total three hierarchical levels of candidate regulators for the dhurrin biosynthesis genes.

5.7. Statistical Analysis

Analysis of growth and chemistry data was performed with GraphPad v.9.0.1. Two‐way ANOVA was used to compare treatments and time points within organ types, with Tukey's or Bonferroni's tests used for multiple comparisons where significance thresholds were met. Data were log₂‐transformed where required to meet assumptions of normality and homoscedasticity.

Author Contributions

A.A.Q., R.M.G., J.E.M., and C.K.B. conceptualized and planned the experiment. A.A.Q. and C.K.B. conducted the plant growth and harvesting. A.AQ. performed chemical analyses and RNA extractions. B.A.M. and A.A.Q. conducted bioinformatic analyses. A.A.Q. performed statistical analyses, interpreted results and prepared original manuscript. All authors contributed to writing and revisions of final manuscript. R.M.G. and J.E.M. secured funding to support the work and supervised the research.

Funding

This research was supported by the Australian Research Council Discovery Project DP210103010 to RMG. AAQ was supported by an Australian Government Research Training Program (RTP) Scholarship and an AW Howard Memorial Trust Research Fellowship. This work was funded in part by the DOE Great Lakes Bioenergy Research Center (DOE BER Office of Science grant/award: DE‐SC0018409) and the Perry Adkisson Chair in Agricultural Biology. The work conducted by the US Department of Energy Joint Genome Institute, a DOE Office of Science User Facility, is supported by the Office of Science of the U.S. Department of Energy under contract no. DE‐AC02‐05CH11231.

Conflicts of Interest

The authors declare no conflicts of interest.

Peer Review

The peer review history for this article is available in the Supporting Information for this article.

Supporting information

Data S1: Peer review.

Table S1: Differential gene expression results for leaf tissue. Contrasts made between nitrogen‐treated (N) and control (C) samples within each timpepoint. Results filtered at false discovery rate (FDR) < 0.05, and fold change (FC) ≥ 5, transcripts per million (TPM) ≥ 5 in at least one sample.

Table S2: Differential gene expression results for stem tissue. Contrasts made between nitrogen‐treated (N) and control (C) samples within each timpepoint. Results filtered at false discovery rate (FDR) < 0.05, and fold change (FC) ≥ 5, transcripts per million (TPM) ≥ 5 in at least one sample.

Table S3: Differential gene expression results for root tissue. Contrasts made between nitrogen‐treated (N) and control (C) samples within each timpepoint. Results filtered at false discovery rate (FDR) < 0.05, and fold change (FC) ≥ 5, transcripts per million (TPM) ≥ 5 in at least one sample.

Table S4: Gene ontology (GO) term enrichment analysis results for leaf. Genes were tested for biological process within time point and fold change (FC) direction for nitrogen‐treated samples relative to controls. GO terms containing an overlap of genes greater than 40%, as determined with Jaccard similarity (JS) index, were grouped. Groups were manually annotated with a representative term (JS manual annotation).

Table S5: Gene ontology (GO) term enrichment analysis results for stem. Genes were tested for biological process within time point and fold change (FC) direction for nitrogen‐treated samples relative to controls. GO terms containing an overlap of genes greater than 40%, as determined with Jaccard similarity (JS) index, were grouped. Groups were manually annotated with a representative term (JS manual annotation).

Table S6: Gene ontology (GO) term enrichment analysis results for roots. Genes were tested for Biological Process within time point and Fold Change (FC) direction for nitrogen‐treated samples relative to controls. GO terms containing an overlap of genes greater than 40%, as determined with Jaccard similarity (JS) index (JS), were grouped. Groups were manually annotated with a representative term (JS manual annotation).

Table S7: Node information for leaf GRNs.

Table S8: Edge table, 24‐h leaf GRN. MR = mutual rank.

Table S9: Edge table, 36‐h leaf GRN. MR = mutual rank.

Table S10: Edge table, 48‐h leaf GRN. MR = mutual rank.

Table S11: Edge table, 60‐h leaf GRN. MR = mutual rank.

Table S12: Transcription factors (TF) in 24‐h leaf GRN. GRN input lists the differentially expressed (DE) TFs present in the GRN input data. GRN output and filter list the TFs remaining in the network after filtering applied in the GRN construction and at visualization. MR = mutual rank, PCC = Pearson's correlation coefficient.

Table S13: Transcription factors (TF) in 36‐h leaf GRN. GRN input lists the differentially expressed (DE) TFs present in the GRN input data. GRN output and filter list the TFs remaining in the network after filtering applied in the GRN construction and at visualization. MR = mutual rank, PCC = Pearson's correlation coefficient.

Table S14: Transcription factors (TF) in 48‐h leaf GRN. GRN input lists the differentially expressed (DE) TFs present in the GRN input data. GRN output and filter list the TFs remaining in the network after filtering applied in the GRN construction and at visualization. MR = mutual rank, PCC = Pearson's correlation coefficient.

Table S15: Transcription factors (TF) in 60‐h Leaf GRN. GRN input lists the differentially expressed (DE) TFs present in the GRN input data. GRN output and filter list the TFs remaining in the network after filtering applied in the GRN construction and at visualization. MR = mutual rank, PCC = Pearson's correlation coefficient.

Table S16: Transcription factors (TF) in 12‐h root GRN. GRN input lists the differentially expressed (DE) TFs present in the GRN input data. GRN output and filter list the TFs remaining in the network after filtering applied in the GRN construction and at visualization. MR = mutual rank, PCC = Pearson's correlation coefficient.

Figure S1: Impact of nitrogen limitation and resupply on morphology of Sorghum bicolor grown under nitrogen limiting conditions and treated with either 12‐mM nitrogen solution (left, nitrogen‐treated) or 0‐mM nitrogen solution (right, control). Image taken prior to the harvest 60‐h after application of treatment.

Figure S2: Experimental setup for time series analysis of response to nitrogen treatment. Sorghum bicolor BTx623 plants were grown in greenhouse conditions under nitrogen limitation (0.5 mM N) until the 6 (± 1)‐leaf stage. A set of control plants were harvested 24 h prior to treatment (baseline harvest). Plants were split into nitrogen treatment or control groups and received a solution containing 12 or a 0 mM N of control at 8 a.m. on the first day of sampling. Plants were harvested at seven intervals over the following 60 h for chemical and RNAseq analysis. (b) Schematic showing the sections of organs harvested for analysis (leaf blade, leaf sheath, stem, and roots). Yellow boxes indicate the sections harvested for transcriptomic analysis (i.e., the midsection of the 6th fully emerged leaf, excluding the midrib, stem, and root tips).

Figure S3: Morphology and growth of Sorghum bicolor plants after delivery of a 12 mM nitrogen treatment or a 0 mM N control. (a) Height (n = 9). (b) Leaf number (n = 9). (c) Total dry mass of leaf blade, leaf sheath and roots (n = 6). (d) Root:shoot ratio (n = 6). Columns represent mean ±SE. Data were analyzed using a two‐way ANOVA. No significant differences were detected between control and nitrogen treatment groups for any growth parameters.

Figure S4: Transcript filtering and principal component analysis (PCA) of RNAseq replicates based on gene expression profiles. Boxplots show expression distributions for individual samples within (a) leaf, (b) stem, and (c) root data, after TMM normalization and filtering to exclude lowly expressed genes. (d) PCA was performed on log₂‐transformed count values for all samples (n = 3, except root 6‐h control and root 60‐h control where n = 2) after filtering and TMM normalization. PC1 and PC2 represents the top two components of variation. Color indicates organ type (leaf—green; root tips—purple; stem—orange), while treatment is represented by shape (control—circles; nitrogen—triangles). Numbers in the legend are the number of hours after treatment, with the depth of shade increasing from light to dark with increasing time.

Figure S5: Volcano plots showing differentially expressed transcripts in nitrogen‐treated samples relative to control samples within each time point for leaf, stem, and root tips. X‐axis represents the log₂ fold change (FC); Y‐axis represents the ‐log10 adjusted p‐value. Transcripts that pass the threshold of|log₂FC| > 2 and FDR < 0.05 are considered differentially expressed and represented as red for upregulated transcripts and blue for downregulated transcripts. The 36‐h root condition was excluded from the analysis due to insufficient replicates.

Figure S6: Counts of gene number in enriched gene ontology (GO) terms for leaf tissue across time points. GO terms were retrieved for lists of upregulated and downregulated genes separately at each time point (“biological process,” p < 0.05, FDR < 0.1). Within each time point, GO terms with overlapping genes were grouped according to Jaccard Similarity scores (0.4). Where GO terms did not form a group, GO IDs are listed alongside the GO Term. Red indicates GO terms enriched in upregulated gene lists. Blue indicates GO terms enriched in downregulated gene lists. Bubble size represents the count of unique genes included in each group.

Figure S7: Counts of gene number in enriched gene ontology (GO) terms for stem DEGs across time points. GO terms were retrieved for lists of upregulated and downregulated genes separately at each time point (“biological process,” p < 0.05, FDR < 0.1). Within each time point, GO terms with overlapping genes were grouped according to Jaccard similarity scores (0.4). Where GO terms did not form a group, GO IDs are listed alongside the GO Term. Red indicates GO terms enriched in upregulated gene lists. Blue indicates GO terms enriched in downregulated gene lists. Bubble size represents the count of unique genes included in each group with a count of > 5 genes.

Figure S8: Counts of gene number in enriched gene ontology (GO) terms for root tissue across time points. GO terms were retrieved for lists of upregulated and downregulated genes separately at each time point (“biological process,” p < 0.05, FDR < 0.1). Within each time point, GO terms with overlapping genes were grouped according to Jaccard similarity scores (0.4). Where GO terms did not form a group, GO IDs are listed alongside the GO term. Red indicates GO terms enriched in upregulated gene lists. Blue indicates GO terms enriched in downregulated gene lists. Bubble size represents the count of unique genes included in each group with a count of > 5 genes.

Figure S9: Gene regulatory network for root tissue 12 h after nitrogen application. Each node represents a transcript which is regulated by a transcription factor (red diamonds). Edges between transcription factors and targe genes (circles) are represented by gray lines. Node size indicates the number of edges (connections) of that node, with larger size indicating greater influence in the network. Genes have been colored according to their function (mapman process). Genes and transcription factors with recognized nitrogen‐related functions have been annotated with their Arabidopsis thaliana gene symbol.

Figure S10: Heatmap of differentially expressed transcription factors in root tips across time and treatment. Heatmaps constructed with means of log₂‐transformed TPM values, further normalized with a z‐score transformation. Clustering by Euclidean distance with complete linkage method for rows (transcripts) and indicated by dendrogram on left of plot. Color scale represents z‐score normalized mean‐log₂(TPM), with red indicating high and blue indicating low expression values. Transcript names for Sorghum bicolor genes are shown to the right of each row. Where known, transcription factor family is indicated in brackets. Transcripts with previously identified functions in nitrogen response are indicated with red text.

Figure S11: Heatmap of differentially expressed transcription factors in stem across time and treatment. Heatmaps constructed with means of log₂‐transformed TPM values, further normalized with a z‐score transformation. Clustering by Euclidean distance with complete linkage method for rows (transcripts) and indicated by dendrogram on left of plot. Color scale represents z‐score normalized mean‐log₂(TPM), with red indicating high and blue indicating low expression values. Transcript names for Sorghum bicolor genes are shown to the right of each row. Where known, transcription factor family is indicated in brackets. Transcripts with previously identified functions in nitrogen response are indicated with red text.

Figure S12: Heatmap of differentially expressed transcription factors in leaf across time and treatment. Heatmaps constructed with means of log₂‐transformed TPM values, further normalized with a z‐score transformation. Clustering by Euclidean distance with complete linkage method for rows (transcripts) and indicated by dendrogram on left of plot. Color scale represents z‐score normalized mean‐log₂(TPM), with red indicating high and blue indicating low expression values. Transcript names for Sorghum bicolor genes are shown to the right of each row. Where known, transcription factor family is indicated in brackets. Transcripts with previously identified functions in nitrogen response are indicated with red text.

Figure S13: Gene regulatory network (GRN) for leaf tissue, 24‐h after nitrogen treatment. Constructed from differentially expressed genes filtered at TPM ≥ 5, FC ≥ |5|, FDR ≥ 0.05. Edges filtered at MR > 0.7. Regulatory relationships between transcription factors (red diamonds) and targe genes (circles) are represented by gray lines (edges). Node size indicates the number of edges (connections) of that node, with larger size indicating greater influence in the network. Genes have been colored according to their function according to Mapman annotation. Dhurrin metabolism genes present in the network are indicated with bold. Arabidopsis thaliana orthologues for known nitrogen‐related enzymes, transporters, and transcription factors are also labeled within the network.

Figure S14: Gene regulatory network (GRN) for leaf tissue, 36‐h after nitrogen treatment. Constructed from differentially expressed genes filtered at TPM ≥ 5, FC ≥ |5|, FDR ≥ 0.05. Edges filtered at MR > 0.7. Regulatory relationships between transcription factors (red diamonds) and targe genes (circles) are represented by gray lines (edges). Node size indicates the number of edges (connections) of that node, with larger size indicating greater influence in the network. Genes have been colored according to their function according to Mapman annotation. Dhurrin metabolism genes present in the network are indicated with bold. Arabidopsis thaliana orthologues for known nitrogen‐related enzymes, transporters and transcription factors are also labeled within the network.

Figure S15: Gene regulatory network (GRN) for leaf tissue, 48‐h after nitrogen treatment. Constructed from differentially expressed genes filtered at TPM ≥ 5, FC ≥ |5|, FDR ≥ 0.05. Edges filtered at MR > 0.7. Regulatory relationships between transcription factors (red diamonds) and targe genes (circles) are represented by gray lines (edges). Node size indicates the number of edges (connections) of that node, with larger size indicating greater influence in the network. Genes have been colored according to their function according to Mapman annotation. Dhurrin metabolism genes present in the network are indicated with bold. Arabidopsis thaliana orthologues for known nitrogen‐related enzymes, transporters and transcription factors are also labeled within the network.

Figure S16: Gene regulatory network (GRN) for leaf tissue, 60‐h after nitrogen treatment. Constructed from differentially expressed genes filtered at TPM ≥ 5, FC ≥ |5|, FDR ≥ 0.05. Edges filtered at MR > 0.7. Regulatory relationships between transcription factors (red diamonds) and targe genes (circles) are represented by gray lines (edges). Node size indicates the number of edges (connections) of that node, with larger size indicating greater influence in the network. Genes have been colored according to their function according to Mapman annotation. Dhurrin metabolism genes present in the network are indicated with bold. Arabidopsis thaliana orthologues for known nitrogen‐related enzymes, transporters and transcription factors are also labeled within the network.

Figure S17: Subnetwork of gene regulatory network (GRN) for leaf tissue, 24‐h after nitrogen treatment. Constructed from differentially expressed (DE) genes filtered at TPM ≥ 5, FC ≥ |5|, FDR < 0.05. Edges filtered at MR > 0.7. Expression data from all time points was included in the analysis. Diamonds = transcription factors. Node color scale = log₂(FC), with red indicating upregulation relative to control and blue indicating downregulation relative to control. Edge color indicated the sign of the correlation (Pearson's correlation coefficient), with red indicating a positive relationship between nodes and blue indicating a negative relationship between nodes. Inferred transcriptional activation of target nodes by regulators is indicated by an arrowhead, while repression of target nodes is indicated by a bar at the edge of the connecting edge. Thick connecting edges between source and target nodes indicated an active regulatory relationship in nitrogen‐treated plants in the GRN for the current time point, while thin lines indicate an inactive regulatory relationship in the nitrogen‐treated plants (but would be active in the control plants). Labels for transcription factors refer to gene symbols for Arabidopsis thaliana orthologues, while gene IDs refer Sorghum bicolor genes.

Figure S18: Subnetwork of gene regulatory network (GRN) for leaf tissue, 36‐h after nitrogen treatment. Constructed from differentially expressed (DE) genes filtered at TPM ≥ 5, FC ≥ |5|, FDR < 0.05. Edges filtered at MR > 0.7. Expression data from all time points was included in the analysis. Diamonds = transcription factors. Node color scale = log₂(FC), with red indicating upregulation relative to control and blue indicating downregulation relative to control. Edge color indicated the sign of the correlation (Pearson's correlation coefficient), with red indicating a positive relationship between nodes and blue indicating a negative relationship between nodes. Inferred transcriptional activation of target nodes by regulators is indicated by an arrowhead, while repression of target nodes is indicated by a bar at the edge of the connecting edge. Thick connecting edges between source and target nodes indicated an active regulatory relationship in nitrogen‐treated plants in the GRN for the current time point, while thin lines indicate an inactive regulatory relationship in the nitrogen‐treated plants (but would be active in the control plants). Labels for transcription factors refer to gene symbols for Arabidopsis thaliana orthologues, while gene IDs refer Sorghum bicolor genes.

Figure S19: Subnetwork of gene regulatory network (GRN) for leaf tissue, 48‐h after nitrogen treatment. Constructed from differentially expressed (DE) genes filtered at TPM ≥ 5, FC ≥ |5|, FDR < 0.05. Edges filtered at MR > 0.7. Expression data from all time points was included in the analysis. Diamonds = transcription factors. Node color scale = log₂(FC), with red indicating upregulation relative to control and blue indicating downregulation relative to control. Edge color indicated the sign of the correlation (Pearson's correlation coefficient), with red indicating a positive relationship between nodess and blue indicating a negative relationship between nodes. Inferred transcriptional activation of target nodes by regulators is indicated by an arrowhead, while repression of target nodes is indicated by a bar at the edge of the connecting edge. Thick connecting edges between source and target nodes indicated an active regulatory relationship in nitrogen‐treated plants in the GRN for the current time point, while thin lines indicate an inactive regulatory relationship in the nitrogen‐treated plants (but would be active in the control plants). Labels for transcription factors refer to gene symbols for Arabidopsis thaliana orthologues, while gene IDs refer Sorghum bicolor genes.

Figure S20: Subnetwork of gene regulatory network (GRN) for leaf tissue, 60‐h after nitrogen treatment. Constructed from differentially expressed (DE) genes filtered at TPM ≥ 5, FC ≥ |5|, FDR < 0.05. Edges filtered at MR > 0.7. Expression data from all time points was included in the analysis. Diamonds = transcription factors. Node color scale = log₂(FC), with red indicating upregulation relative to control and blue indicating downregulation relative to control. Edge color indicated the sign of the correlation (Pearson's correlation coefficient), with red indicating a positive relationship between nodes and blue indicating a negative relationship between nodes. Inferred transcriptional activation of target nodes by regulators is indicated by an arrowhead, while repression of target nodes is indicated by a bar at the edge of the connecting edge. Thick connecting edges between source and target nodes indicated an active regulatory relationship in nitrogen‐treated plants in the GRN for the current time point, while thin lines indicate an inactive regulatory relationship in the nitrogen‐treated plants (but would be active in the control plants). Labels for transcription factors refer to gene symbols for Arabidopsis thaliana orthologues, while gene IDs refer Sorghum bicolor genes.

Data Availability Statement

The datasets analyzed during the current study are available through the JGI Genome portal https://genome.jgi.doe.gov/portal through the projects: Sorghum bicolor BTx623 Nitrogen Grown Conditions Gene Expression Profiling (project IDs: 1404259 and 1,456,498). For additional assistance on how to access the data, the corresponding author may be contacted.