To investigate the response of rice to different nitrogen forms, we used a hydroponic system in which ammonium and/or nitrate could be supplemented to the medium. 5mM of nitrate (NO₃ ^- as KNO₃), 5mM of ammonium (NH₄ ⁺ as (NH₄)₂SO₄), an equimolar combination of both nitrogen forms (2.5mM of NH₄NO₃) or a control (5mM K⁺ as K₂SO₄) with potassium (K⁺) balanced at 5mM among all treatments as K₂SO₄, were supplemented into the nitrogen-free growing media of the rice seedlings 5 days after germination and the seedlings were let grown for 10 more days before phenotyping (see Materials & Methods for details on the procedure). In our set-up, supplementation with ammonium and nitrate had a similar positive effect on shoot biomass, while co-application of both forms showed a synergistic positive effect ( Figure 1A ). The lateral root density positively correlated with the shoot biomass and showed a similar synergistic response to the combined treatment. The root system treated with nitrate had a long primary root with long lateral roots close to the root-hypocotyl junction, while the ammonium-supplemented root system had a dense network of small lateral roots evenly spread over the primary root ( Figures 1B–E ). Co-application seemed to result in a combination of the two phenotypes. Finally, we observed an increase in leaf chlorophyll content upon treatment by ammonium or ammonium-nitrate but not by nitrate alone ( Figure 1F ).

We used the same hydroponic system as described above to collect samples for the transcriptomic analysis, but rice tissues were harvested soon after the nitrogen supplementation (see Material and Methods for details). In Arabidopsis, early response genes are induced as early as 12 minutes (NITRITE REDUCTASE1 (NIR1)), 15 minutes (NRT2.1 and NITRATE REDUCTASE1 (NIA1)) or 20 minutes (NITRATE TRANSPORTER1.1 (NRT1.1)) after nitrate treatment (Krouk et al., 2010). Therefore, to capture relevant transcriptional profiles, we sampled root and shoot tissue separately immediately (0h), 15 minutes, 1h, 2h, 4h, 12h, 24h and 48h after treatment and used these samples for RNA sequencing (RNA-seq) thereby generating an extensive dataset covering the nitrogen transcriptional responses in rice ( Figure 2 ).

We performed a pair-wise differential analysis to assess differential expression for each time point and treatment ( Supplementary Dataset S1 , Supplementary Dataset S2 ). Considering an absolute fold-change >2, and an adjusted p.value (FDR) < 0.05, a significant number of genes were differentially expressed by the treatments in the shoot or root and over the time-course ( Supplementary Figures S1 – S3 ; Supplementary Dataset S1 , Supplementary Dataset S2 ). Nitrate, alone or in combination with ammonium, rapidly induced over 250 genes in the roots within 15 minutes ( Supplementary Figures S1 ). This list includes homologues of Arabidopsis primary nitrate response genes such as LOB DOMAIN-CONTAINING PROTEIN37/38/39 (LBD37/38/39), NITRATE-INDUCIBLE GARP-TYPE TRANSCRIPTIONAL REPRESSOR1 (NIGT1), NRT1.1, nitrate and nitrite reductases, GLUCOSE-6-PHOSPHATE DEHYDROGENASE3 (G6PDH3) and ARABIDOPSIS NAC DOMAIN-CONTAINING protein 4 (NAC4) ( Supplementary Dataset S1 ). In contrast, the response to ammonium was very weak at the 15 minutes-timepoint but a high number of differentially expressed genes was observed after 1h ( Supplementary Figures S1 , S3 ; Supplementary Dataset S1 ), including the transporter-encoding AMMONIUM TRANSPORTER1.2 (OsAMT1.2) and OsAMT2.2 or the amino acid assimilation enzyme-encoding ALANINE AMINOTRANSFERASE1 (OsAlaAT1), OsAlaAT2, ASPARAGINE SYNTHETASE1 (OsASN1), PHOSPHOENOL PYRUVATE CARBOXYKINASE1 (OsPPCK1), and GLUTAMATESYNTHASE1 (OsGLT1). The highest number of differentially expressed genes was in general observed with the combined treatment of ammonium and nitrate. The majority of these genes were also affected by either the ammonium or nitrate treatment ( Supplementary Figure S1 ). Hence, the combined ammonium-nitrate response seems to largely reflect the sum of the individual responses.

In the shoot, a strong response only occurred from 4h onwards, primarily attributable to the nitrate treatment. The ammonium treatment resulted in a slower response, but from 12h onwards, large transcriptomic changes were observed as well ( Supplementary Dataset S2 , Supplementary Figure S2 ).

To analyze the gene response profiles towards the different treatments, we built a co-expression network using the R package WGCNA (Langfelder and Horvath, 2008) for the most varying genes in the roots (18457) and shoots (18343). The network revealed 54 co-expression clusters in the roots and 55 in the shoots ( Supplementary Figures S4 , S5 ; Supplementary Datasets S1 , S2 ). The accompanying edge and node tables, compatible with network visualization tools such as Cytoscape or Gephi can be downloaded at https://osf.io/2uzd3/. To provide access to these resources, we generated a Shiny app Supplementary Figure S6 ), https://www.psb.ugent.be/shiny/rice-response-to-nitrogen/). The user can query any of the 42189 rice genes to display the expression profile in response to the different nitrogen treatments. If the gene is also included in the 18457 genes or 18343 genes used for the co-expression network, the Eigengene of its WGNCA cluster and a correlation coefficient with highly correlated genes (biweight midcorrelation, computed during the gene co-expression network creation) is also displayed. The latter is also shown in Supplementary Datasets S3 and Supplementary Dataset S4 , which facilitate the identification of highly co-expressed gene pairs. The cluster membership and associated p values indicating the contribution to the cluster profile for each gene as well as the number of connections to other genes within the same cluster are indicated in Supplementary Datasets S1 , S2.

In the roots, we identified clusters specifically and early induced by nitrate (nitrate and ammonium-nitrate treatments only) containing transiently (‘green3’) or constitutively induced genes (‘thistle3’) ( Figure 3 ; Supplementary Figure S4 ). We identified two clusters specifically induced by ammonium (‘darkslateblue’ and ‘deeppink1’). Two clusters of genes were induced by ammonium and with an approximately 4h delay by nitrate or weaker induction by nitrate, possibly due to the nitrate to ammonium reduction (‘mediumorchid’, ‘thistle4’). We identified small clusters with a specific response to ammonium (‘yellow3’) or nitrate (‘indianred3’), but no or very weak response to the combination of the two nitrogen forms, indicative for a countereffect of the other nitrogen form on these genes. Vice versa, we did not identify clusters of genes induced by the ammonium-nitrate treatment only. Some other clusters show a similar response to all nitrogen forms, and are likely related to the nitrogen nutrition. Most other clusters showed a high response in the mock as well or show irregular or variable expression profiles ( Supplementary Figure S4 ).

In the shoots, we identified early responsive and nitrate-specific clusters that are similar to the nitrate-specific clusters in the roots, including a transient (‘pink2’, similar to ‘green3’) and a constitutive cluster of upregulated genes (‘lightcyan1’, similar to ‘thistle3’). Also a cluster of genes exclusively induced by nitrate could be observed (‘plum4’), similar as the ‘indianred3’ cluster in the roots. Contrary to the roots, we did not identify an ammonium-specific cluster in the shoots. Moreover, many more shoot clusters exhibit irregular patterns or show similar responses in the mock as in the treatments, making them of less interest. Overall, our co-expression networks revealed clusters of genes illustrating strong temporal and differential biological responses to the different forms of nitrogen provided.

To further investigate the clusters nature, we conducted a gene-ontology enrichment analysis ( Supplementary Datasets S5 , S6 ). We first compared the nitrate-specific clusters in the roots (‘green3’, ‘thistle3’) and the shoots (‘pink2’, ‘lightcyan1’). The genes ontologies enriched in both roots and shoots nitrate-specific clusters are highly similar and many genes are retrieved in both clusters: 72.8% of the 125 genes composing the nitrate-specific shoot clusters are retrieved in the 414 genes composing the nitrate-specific root clusters. The genes present in all these clusters are primarily related to nitrate assimilation and nitrate transport.

Highly enriched terms for the ammonium-specific clusters in the roots ‘darkslateblue’ and ‘deeppink1’ are mainly related to ammonium or amino acid assimilation and cellular respiration or ATP production. The ‘indianred3’ root cluster and the ‘plum4’ shoot cluster, containing genes that are exclusively induced by nitrate alone, are both highly enriched in iron-related terms. The root cluster ‘yellow3’ showing an exclusive response to ammonium alone, mainly concerns genes related to oxidative stress ( Supplementary Datasets S3 , S4 ).

For further analysis of the co-expression network, we zoomed in on the two main nitrate-specific clusters in the root network (‘green3’ and ‘thistle3’) containing genes that were rapidly induced upon nitrate ( Figure 4 ; Supplementary Table S1 ). With for example OsNRT1.1B (Os10g40600) and NITRATE REDUCTASE1 (OsNR1) (Os08g36480), this group contains typical nitrate sentinel genes. In the same group, we identified 38 transcription factors based on PlantTFDB v5.0 (https://planttfdb.gao-lab.org/) (Tian et al., 2019) ( Supplementary Dataset S1 ). Several of these transcription factors have a high module membership and a high number of connections within one of the two clusters and could be designated as ‘hub’ genes with potentially an important role in the nitrate response or signaling ( Figure 4 , Table 1 ). A highly connected transcription factor in ‘green3’ is OsLBD38 (Os03g41330) which homologues were shown to be involved in nitrogen signaling in Arabidopsis or other species (Rubin et al., 2009; Teng et al., 2022), while OsLBD38 seems to be part of a conserved regulatory cluster between Arabidopsis and rice (Obertello et al., 2015). OsLBD38 is also the most connected transcription factor in the shoot cluster ‘lightcyan1’ ( Supplementary Dataset S2 ). OsNIGT1 (Os02g22020), known to be an important transcriptional regulator of the nitrate signaling as well, is also present in ‘thistle3’ ( Figure 4 , Table 1 ; Supplementary Dataset S1 ) (Maeda et al., 2018). Several transcription factors have come forward that have not been previously related to nitrate response. OsGRAS49 (Os11g47890) for instance, which is to our knowledge not reported to have a role in the nitrate response, is a potential ‘hub’ transcription factor in the nitrate specific clusters in both roots and shoots ( Supplementary Dataset S1 , S2 ).

In this study, we provided a detailed overview of the transcriptional response of rice in roots and shoots to different nitrogen forms and generated a resource with the expression profile of any rice gene of interest in response to nitrate, ammonium, or the combination of both (all expressions profiles are available on https://www.psb.ugent.be/shiny/rice-response-to-nitrogen/). We used this dataset to generate a co-expression network, and identified clusters with a specific response to nitrogen in both roots and shoots. Furthermore, the co-expression network created the possibility to infer putative transcription factors/target genes relationships. As the different nitrogen treatments lead to distinct variations due to unique interactions, we anticipated uncovering otherwise overlooked regulatory relationships. We illustrated this by identifying new transcription factors with a role in nitrate signaling and showing the potential effect of two transcription factors, OsRLI1 and OsEIL1, on the induction of nitrate response.

OsRLI1 is a transcription factor involved in phosphate starvation signaling (Zhang et al., 2021). As a matter of fact, nitrate is known to affect the phosphate signaling pathway (Hu et al., 2019). Supporting this, our co-expression network revealed that OsRLI1 expression is correlated with the expression of several phosphate-starvation signature induced genes: INOSITOL-3-PHOSPHATE SYNTHASE ISOZYME1 (OsIPS1) (often used as a phosphate starvation reporter (Hou et al., 2005; Wang et al., 2009a; Dai et al., 2012)), 1-AMINOCYCLOPROPANE-1-CARBOXYLIC ACID SYNTHASE (OsACS) [involved in tolerance to phosphate starvation in rice (Lee et al., 2019)], SPX-MAJOR FACILITY SUPERFAMILY2 (OsSPX-MSF2) (involved in phosphate signaling/transport and induced by phosphate starvation (Wang et al., 2012)), and finally PHOSPHATE STARVATION RESPONSE2 (OsPHR2) which is the main regulator of phosphate starvation responses (Zhou et al., 2008) and inducer of OsRLI1 (Zhou et al., 2008; Wu and Wang, 2011; Zhang et al., 2021). OsRLI1 is moreover a close homologue of OsPHR2 and AtPHR1 and interacts just as these with SPX-domain containing proteins (Puga et al., 2014; Wang et al., 2014; Ruan et al., 2018; Zhang et al., 2021). OsPHR2 binds to OsSPX1/2/4 upon high phosphate. A low cellular inositol phosphate level, which depends on the phosphate level of the cell, disrupts the SPX retention of OsPHR2 which then is free to migrate to the nucleus where it binds to phosphate starvation inducible genes promoters (Wild et al., 2016; Crombez et al., 2019; Hu et al., 2019). At least the interaction with OsSPX4 depends also on nitrate levels: the transceptor OsNRT1.1B can promote OsSPX4 protein degradation in a nitrate-dependent manner, impacting directly the phosphate signaling pathway (Hu et al., 2019). OsRLI1 was shown to be induced by nitrate to induce the phosphate starvation response and finetune the N-P balance (Zhang et al., 2021). Our results show that it may also induce nitrate responsive genes, further complicating the phosphate-nitrate crosstalk.

OsEIL1 is a transcription factor involved in ethylene signaling (Yang et al., 2015a, Yang et al., 2015b) and regulates various genes such as transcription factors and metabolic genes (Dolgikh et al., 2019) and hormonal pathways (Chang et al., 2013). Here, we showed that OsEIL1 upregulation by nitrate correlates with OsNRT1.1B induction in rice. In Arabidopsis, nitrate induces ethylene production via induction of 1‐aminocyclopropane‐1‐carboxylic acid (ACC) synthases (ACS) and ACC oxidases (ACO), key enzymes in the ethylene biosynthesis pathway (Kende, 1993; Khan et al., 2015). Moreover, nitrate-induced expression of NRT1.1 requires ethylene signaling (Tian et al., 2009), but it is not known how these pathways exactly connect to each other. Additionally, certain nitrate transporters were shown to be directly controlled by ethylene (Zheng et al., 2013; Zhang et al., 2014). As in Arabidopsis, multiple ACS genes are in our dataset induced upon nitrate in our rice dataset, including OsACS2, OsACS5 and OsACS6, supporting a comparable pathway in rice and Arabidopsis. However, the absence of binding motifs for the OsEIL1 transcription factor (Hiraga et al., 2009) or ethylene response factors ERFs (Ohme-Takagi and Shinshi, 1995) in the promoters of OsNRT1.1B and OsNR1 argue for an indirect impact on these genes by OsEIL1. Still, our results show that OsEIL1 is not only able to – possibly indirectly – induce OsNRT1.1B and OsNR1, but also that OsEIL1 is important for the nitrate-induced expression of those genes, featuring OsEIL1 as a central transcription factor in the ethylene signaling-dependent nitrate response in rice.

While we focused on the nitrate-specific clusters to investigate new candidate regulators, other parts of the co-expression network can be explored as well. For instance, the ammonium-specific cluster may provide valuable insights into ammonium signaling, although this could be more challenging due to the generally slower transcriptional response compared to nitrate. This slow transcriptional response indicates that ammonium does not directly activate a transcriptionally regulated signaling pathway. Still, ammonium is suggested to be signaling molecule (Liu and von Wirén, 2017). Bacteria have even been shown to possess an ammonium-sensing histidine kinase (Pflüger et al., 2018, Pflüger et al., 2024), but similar mechanisms have not yet been demonstrated in plants. Interestingly, the bacterial sensor is part of the ammonium transporter/methylamine permease/Rhesus family, which also includes plant AMT proteins that have been proposed to function as ammonium receptors (Liu and von Wirén, 2017). The fact that ammonium does not induce rapid transcriptional changes in rice does not exclude that ammonium can act as a potential signaling molecule via another biochemical pathway and indirectly trigger a transcriptional response. In this respect, it is important to note that we observed a strong transcriptional response starting 1 hour after treatment, with a considerable number of transcription factors identified in the ammonium-specific clusters, including for example MONOCULM1 (OsMOC1, Os04g35250), OsNAC5 (Os11g08210) and OsNLP6 (Os02g04340) that showed high expression levels (FC > 8) after 1 hour of treatment. Interestingly, OsNLP6 is a homolog of OsNLP1, OsNLP3, and OsNLP4 which are all known for their implication in nitrate and ammonium responses or in nitrogen use efficiency (Hu et al., 2019; Alfatih et al., 2020; Wang et al., 2021; Wu et al., 2021). OsNLP6 is only known for having a very low basal expression and not responding to various stress tested in previous studies, but was never characterized further (Jagadhesan et al., 2020; Wu et al., 2021). The high expression of OsMOC1 is somewhat surprising as it is mainly known for its critical role in regulating tiller number and plant architecture (Liao et al., 2019). Finally, OsNAC5 is an abiotic stress-responsive gene (Takasaki et al., 2010), which might indicate that a stress induce the transcriptional response. Ammonium is known to affect rapidly the internal and external pH of roots, which may be the chemical cue resulting in this response (Jia et al., 2020; Motte and Beeckman, 2020). We also observed that ammonium upregulated alanine aminotransferases expression, indicating an accumulation of alanine in planta. Such responses are usually observed in stress conditions to store nitrogen and to provide energy and reductants under for instance anoxia situations in the cell (Vanlerberghe et al., 1991; Miyashita et al., 2007). Alanine biosynthesis is a known ammonium detoxification process with alanine serving as a nitrogen store (Esteban et al., 2016). In Arabidopsis roots, hypoxia induces AlaAT1 and AlaAT2 as early as 2h after stress application with a peak at 8h, followed by a decrease after 24h, which corresponds to what we and others observed in rice upon ammonium treatment (Miyashita et al., 2007) and was also observed in maize (Muench et al., 1998). Gene ontology enrichment for the ammonium-specific cluster (‘darkslateblue’) revealed an increase in proton related ATPase activity terms potentially indicating a response to counteract cytoplasmic acidification caused by ammonium uptake, thereby contributing to ammonium tolerance in rice. The enrichment of the pyruvate metabolic process term suggests a higher demand for energy production or amino acid biosynthesis, as pyruvate is a central metabolite connecting glycolysis, the TCA cycle, and the amino acid synthesis pathways. Overall, this suggests that the response is more likely related to acidification or stress rather than ammonium acting as a signaling molecule. In any case, the poor overlap in response to ammonium in the shoot and root supports a local effect.

Both in our and previous studies, co-application of ammonium-nitrate resulted in more growth compared to both forms individually ( Figure 1 ) (Kronzucker et al., 1999). Our data showed a broader transcriptional response to the combined nitrogen treatment, encompassing responses that are otherwise only elicited by either ammonium or nitrate alone. This is particularly clear in the cluster analysis, where the ammonium-nitrate profile closely follows either the ammonium or nitrate expression patterns, but rarely exhibits a distinct profile. Hence, the combined provision may elicit a dual action that translate into improved growth. This was specifically observed in lateral root density, where the spatial distribution resulting from the combined treatment resembled the cumulative distribution patterns observed under each individual nitrogen form. Additionally, the ammonium treatment resulted in higher leaf chlorophyll content, which is in line with the positive effect of ammonium on photosynthesis activity as reported in Arabidopsis (Sanchez-Zabala et al., 2015). This effect was also observed with the ammonium-nitrate combination, but not with nitrate alone, further illustrating that the action of one of the forms is preserved within the combined treatment.

An alternative explanation for the differences in growth between co-application and single application is that the provision of only one nitrogen form could trigger a stress response, which is absent when both forms are present. Indeed, despite rice being considered as an ammonium-tolerant plant, we observed that ammonium supplementation alone reduces the size of the rice root system, a phenotype typically associated with ammonium toxicity (Liu and von Wirén, 2017). Accumulation of chlorophyll is in Arabidopsis associated with a mild ammonium stress (Sanchez-Zabala et al., 2015). Likewise, the ‘yellow3’ co-expression cluster that group genes induced by ammonium but not by ammonium-nitrate shows an oxidative stress signature, while a number of stress-related genes are induced upon ammonium treatment (see above). Hence, while considered to be ammonium tolerant, rice clearly displays toxicity-related phenotypes, as also observed in other recent studies (Jia et al., 2020; Xie et al., 2023; Yan et al., 2023). The presumed ammonium tolerance likely originates from observations of paddy field-grown rice, where ammonium is partially converted to nitrate, and rice at the end perceives both ammonium and nitrate. Furthermore, genes in the ‘indianred3’ and ‘plum4’ clusters that are exclusively induced by nitrate only and by none of the other treatments are primarily linked to iron homeostasis and transport as illustrated by the GO enrichment. Such genes, including OsIRO2, OsIRO3, OsNRAMP1, OsPOT, OsOPT7 and OsMIR are typically upregulated upon iron starvation (Zheng et al., 2009; Zhang et al., 2017), which is known to occur when nitrate is the sole nitrogen form provided (Chen et al., 2018). Hence, the observed improvement in growth with the combined treatment may be attributed to the mitigation of stress effects that are typically induced by the individual nitrogen forms.

By focusing on a few nitrate-specific clusters, we demonstrated that our dataset, which includes responses to both ammonium and nitrate, can be utilized to identify candidate transcription factors involved in nitrogen signaling. Other clusters with different nitrogen response profiles presented in this study can be investigated as well, either to identify novel regulators or to predict functions for unknown genes. For instance, uncharacterized putative transporter encoding genes that were identified as strongly co-expressed with nitrate transporters in our network might encode transporters with a role in nitrate transport. Overall, our present study provides the research community with an extensive dataset describing how rice, a major staple crop, responds at the transcriptional level to two main nitrogen feedstocks. A better understanding of how plants sense, take up and process the two main forms of nitrogen provided by fertilization is an important field of study within the contemporary context of the increasing need to breed crop plants with enhanced nitrogen use efficiency.

Rice seedlings [Oryza sativa Nipponbare cultivar (#GSOR100, USDA-ARS)] were dehulled and sterilized with ethanol 70% for 5 minutes, followed by immersion in bleach 6% with Tween-20 for 30 minutes. Seedlings were imbibed by immersion in sterile water for 12h to synchronize germination at 30 degrees. Germinating seeds were transferred on a hydroponic system, and roots were immersed in a nitrogen-free basal salt medium composed of K₂SO₄ 0.7mM, KH₂PO₄ 0.3mM, CaCl₂.2H₂O 1mM, MgSO₄.7H₂O 1mM, Na₂SiO₃.9H₂O, Na₂-Fe-EDTA 20µM for macronutrients, and MnCl₂.4H₂O 9µM, Na₂MoO₄.2H₂O 0.39µM, H₃BO₃ 20µM, ZnSO₄.7H₂O 0.77µM, CuSO₄.5H₂O 0.32µM for micronutrients (pH 5.8). Seedlings were then transferred to a growth cabinet in the dark at 30 degrees for 3 days in a randomized block design. The light was then turned on after 72h and let on for 48h before treatment occurred. Nitrogen treatments consisted of injection with 5mM KNO₃ (5mM NO₃ ^- treatment), 2.5mM (NH₄)₂SO₄ + 2.5mM K₂SO₄ (5mM NH₄ ⁺ treatment), 2.5mM KNO₃ + 1.25mM (NH₄)₂SO₄ + 1.25mM K₂SO₄ (2.5mM NH₄ ⁺ and 2.5mM NO₃ ^- treatment) or 2.5mM K₂SO₄ (mock treatment) in this basal medium. K₂SO₄ was used to balance potassium (K⁺) equimolarly to 5mM in each of the treatments. Rice seedlings were extracted 15min, 1h, 2h, 4h, 12h, 24h and 48h after nitrogen treatments. A supplemental control without treatment was extracted at the 0h time point in 3 biological replicates for roots and shoots, to estimate the impact of the manipulation of the samples (referred to as “Control 0h”). At the extraction time-point, shoots and roots were cut with a razor blade and frozen in liquid nitrogen. The remaining seeds were discarded. Three different boxes were used for each treatment and for each time-point, for a total of 87 boxes. At least 10 germinated seedlings were sampled per box.

For the phenotyping experiments, the same procedure as described above was followed but seedlings were let grown in the hydroponic media for 10 days after treatment and the medium was refreshed daily. Chlorophyll was extracted with DMSO and measured by absorbance at 663nm (Chlorophyll A) and 645nm (Chlorophyll B). Chlorophyll content was measured as:

Frozen roots and shoot samples were grinded with one 3mm metal bead into Eppendorf tubes. RNA was extracted with Trizol (Life Technologies) and the RNeasy Mini Kit (Qiagen) following the manufacturer instructions. An extra DNase step was performed with RNase-Free DNase Set (Qiagen). RNA samples were resuspended in RNAse free water. RNA concentration and purity were determined spectrophotometrically using the Nanodrop ND-1000 (Nanodrop Technologies) and RNA integrity was assessed using a Bioanalyzer 2100 (Agilent).

The sequencing and library preparation was performed by the VIB Nucleomics Core Facility (Leuven, Belgium; www.nucleomics.be). Per sample, 500ng of total RNA was used as input. Using the Illumina TruSeq^® Stranded mRNA Sample Prep Kit (protocol version: Part # 15031047 Rev. E - October 2013), poly-A containing mRNA molecules were purified from the total RNA input using poly-T oligo-attached magnetic beads. In a reverse transcription reaction using random primers, RNA was converted into first strand cDNA and subsequently converted into double-stranded cDNA in a second strand cDNA synthesis reaction using DNA PolymeraseI and RNAse H. The cDNA fragments were extended with a single ‘A’ base to the 3’ ends of the blunt-ended cDNA fragments after which multiple indexing adapters were ligated introducing different barcodes for each sample. Finally, PCR enrichment was conducted to enrich those DNA fragments that have adapter molecules on both ends and to amplify the amount of DNA in the library. Sequence-libraries of each sample were equimolarly pooled and sequenced on Illumina NextSeq 500 (High Output, 75 bp, Single Reads, v2). The raw transcriptomic data (fastq files) have been deposited in the functional genomics data collection ArrayExpress under the accession number E-MTAB-13146.

All analyses were done on the VIB-UGent Plant System Biology Galaxy platform (Afgan et al., 2018). The Trimmomatic tool (Bolger et al., 2014) was used to trim the reads for low-quality read-ends with the following options: raw fastq file, type TrueSeq3 adapter sequences. Data quality was assessed with the FastQC tool before and after trimming with the Trimmomatic tool. The output of Trimmomatic was processed by the Salmon tool (Patro et al., 2017). Salmon was used for transcript-level quantification estimates of RNAseq data. The reads were mapped on the coding sequences of release 7 of the MSU Rice Genome Annotation Project (Kawahara et al., 2013) with the following options: stranded reads and reads derived from the reverse strand, with an Incompatible Prior setting of 1x10^-20. Salmon acts in two steps: the indexation of the reference genome (Oryza sativa japonica v7JGI) and the mapping of the reads trimmed by Trimmomatic to this reference genome, followed by their quantification. The output is an estimated number of reads in transcript per millions. The package txtimport 1.14.0 (Soneson et al., 2015) in the R Statistical software version 3.4.3 was used to process the Salmon output data (transcript-level abundance) and summarize it into matrices of counts of reads/fragments (gene-level abundance).

The gene co-expression network and clusters were built using the WGCNA package (Langfelder and Horvath, 2008). We used the varianceStabilizingTransformation() (VSD) function of the package DESeq2 to transform and normalize the DESeqDataSet data described above without the outliers, as recommended for big experiments containing more than 100 samples, and averaged the 3 biological samples per treatment, per time-point. Only genes with more than 5 counts in at least 2 repetitions per treatment per time point were kept, removing non or very lowly expressed genes. This first threshold reduced the total number of genes to around 26000 for roots and shoots. For computational reasons and to remove noise background, a second threshold removing the 30% least-varying genes based on their expression variance between the treatments as recommended by the WGCNA developers was applied. The final input for the gene co-expression network construction was 18343 genes for the shoots and 18457 genes for the roots. DatasetGene connectivity was determined with a power β (soft thresholding) of 7 for the roots and to 8 for the shoots, chosen with the function pickSoftThreshold() with the following options: networkType = “signed hybrid”, corFnc = “bicor”, maxPOutliers = 0.02. The function ‘adjacency()’ was used with the same options. The options used to design the network with the function cutreeDynamic were deepSplit = 3, and minModuleSize = 20. For every cluster generated, a cluster eigengene is computed; this eigengene (first principal component of a cluster) can be seen as representative of all the genes that compose the cluster. Eigengenes with a correlation with another eigengene higher than 80% (R2 = 0.8) were merged into one cluster. Network visualization was done with Cytoscape 3.7.2 (Shannon et al., 2003)

To identify enriched biological processes, molecular functions, and cellular components within co-expression clusters, a Gene Ontology (GO) enrichment analysis was performed using the GO enrichment tool of the Plaza Monocots 4.0 Platform (Van Bel et al., 2018) using the Locus ID and the publicly available Rice v7.0JGI database with the whole annotated genome as the reference set. The significance threshold for enriched GO terms was set at a p-value of 0.01.

Transcription factor coding sequences were isolated by PCR from rice shoots or root cDNA and used to generate the ‘inducers plasmids’. Promoter sequences of the target genes were isolated from genomic DNA and correspond to the -2000bp sequence upstream of the start codon of the target gene or were limited by the presence of another gene downstream and used to generate the ‘target plasmids’. The plasmids were constructed with the Golden Gateway assembly system: in the inducer plasmids, the coding sequences of the transcription factors were combined with a constitutive promoter (p35s) followed by a nuclear localization sequence. A NOST terminator was placed downstream of the gene coding sequence. In the target plasmids, the genes promoters were cloned upstream of a nuclear localization sequence followed by the fluorescent protein mEGFP coding sequence and a NOST terminator. The inducers plasmids structure can be summarized as “p35s::NLS::transcription-factor-CDS::NOST”. The target plasmids structure can be summarized as “gene-promoter::NLS::mEGFP::NOST”. Sequences were validated by sequencing (Eurofins Genomics, Belgium) and reference sequences were extracted from the Plaza Monocots 4.0 Platform (Van Bel et al., 2018). The list of primers used for the genes coding sequences and promoter isolation is available in Supplementary Table S1 .

14-days old rice seedlings (#GSOR100 USDA-ARS) grown in the dark in sterile vitro-vent boxes on a solid media containing 0.305g/l Murashige & Skoog Modified Basal Salt Mixture Nitrogen-free salts (Phytotech Labs #M407), 0.6mM KH₂PO₄, 9.4mM K₂SO₄, 1mM NH₄NO₃, 1.6mM Na₂SiO₃.9H₂O, 8g/l agar and 0.025g/l MES at pH 5.7, were harvested by cutting the stem above the seed and the aerial part kept for protoplast isolation. The protoplasts extraction and transformation followed the protocol described in other studies with few adaptations (Abel and Theologis, 1994; Yoo et al., 2007; Zhang et al., 2011). Briefly, once extracted, the protoplasts were mixed with different combinations of one inducer plasmid and one target plasmid. Addition of PEG-4000 to the mix induced the transient transformation of the protoplasts which assimilated the different combinations of the two types of plasmids, and transformation was stopped after 15 minutes. After incubation overnight, the protoplasts in solution were distributed in a 90-well plate and mEGFP fluorescence intensity (excitation: 488nm, emission: 522nm) was measured by confocal microscopy.

The OsEIL1 knock-out mutant was generated in a Japonica variety Wuyunjing-7 (9522) using the CRISPR-Cas9 technique, while OsRLI1 knock-out mutant is Japonica variety Nipponbare background and was generated in a previous study (Ruan et al., 2018). Homozygous mutant lines were used for subsequent analysis.

Rice seeds of wild-type and mutant lines were sterilized with 70% (v/v) ethanol for 1 min, followed by 30% (v/v) sodium hypochlorite solution for 30 min. Seedlings were imbibed by immersion in sterile water for 12h to synchronize their germination and let grown in the dark on nitrogen free solution for 3 days, and then transferred to the growth chamber (30 degrees, continuous light) for another 3 days. Seedlings with ~2 cm seminal root were selected for different nitrogen treatments with modified Kimura B solution: high nitrogen (1.5 mM (NH₄ ⁺)₂SO₄, or 3 mM KNO₃ ^-, HN) and nitrogen free (- N or N-free). The time course started at the moment of the transfer. 20 seedlings roots per technical replicate where harvested, and samples were processed as described above for the transcriptome experiment. The RNA was synthetized into cDNA, and the primers presented in Supplementary Table S1 were used for the RT-qPCR as previously described (Xie et al., 2023)

Geminated rice seedlings were first grown in water for 3 days in a growth chamber under a photoperiod of 14 h light (200μmol m-² s-² light density and 70% humidity) and a temperature of 28 degrees, and rice seedlings with ~2 cm long seminal root were then transferred to the hydroponic culture supplied with modified Kimura B solution (500 mL volume for each cup with 10 seedlings) for different nitrogen treatments. For nitrogen -free treatment, nitrogen sources (NH₄)₂SO₄ and KNO₃ was replaced with K₂SO₄ at a concentration of 1.5 mM; for NH₄ ⁺ treatment alone, KNO₃ was replaced with K₂SO₄ at the same concentration; for NO₃ ^- treatment alone, (NH₄)₂SO₄ was replaced with 3 mM KNO₃. The 2-[morpholino]ethane sulfonic acid (MES) was supplied to hydroponic cultures to buffer pH of the medium when mentioned. The rice seedlings were treated for 4 days, and the nutrient solution was renewed every two days.