Introduction

Plants determine the seasonal timing of flowering based on environmental cues such as day length and temperature (1–6). The small protein FLOWERING LOCUS T (FT) is a florigen, a mobile signaling molecule that promotes flowering (7–10). In Arabidopsis thaliana, some phloem companion cells residing in the distal parts of leaves highly express FT (11, 12). Although FT exhibits only a single expression peak in the evening under common laboratory long-day (LD) conditions, FT is highly expressed in the morning and evening under natural light conditions (13, 14). The discrepancy between natural and laboratory conditions can be attributed to differences in the red-to-far-red light (R/FR) ratios. Adjusting the R/FR ratio to that observed in nature is sufficient to recreate the bimodal expression pattern of FT in the lab (13, 14).

Despite extensive research on the genetic regulation of FT expression and function, our understanding of the cells that express FT is limited (6). Although FT is expressed in some phloem companion cells, the precise molecular features that distinguish FT-expressing phloem companion cells from other companion cells have remained elusive (6). Recently, to understand the distinct features of FT-expressing cells, we employed Translating Ribosome Affinity Purification (TRAP)-seq, a method that identifies ribosome-associated transcripts in specific tissues or cell types (15). We found that the list of differentially translated transcripts in FT-expressing cells partially overlapped with that in the general phloem companion cell marker gene SUC2-expressing cells but also contained unique transcripts to FT-expressing cells (15). This supports the notion that the FT-expressing cells are similar but different from general phloem companion cells (11, 12). Our TRAP-seq datasets showed that several FT positive and negative transcriptional regulators and proteins involved in FT protein transport were specifically enriched in FT-expressing cells. We also found that a gene encoding the small protein FPF1-LIKE PROTEIN 1 (FLP1) is highly expressed in FT-expressing cells under LD conditions with adjusted R/FR ratio (hereafter, LD+FR conditions), but not under standard laboratory LD conditions (15). Further analysis revealed that FLP1 promotes flowering and initial inflorescence stem growth, suggesting that it orchestrates flowering and inflorescence stem growth during the transition from the vegetative to the reproductive stage.

Single-cell RNA-seq (scRNA-seq) is a powerful tool to study different cell populations within the complexity of biological tissues. Although our bulk tissue/cell-specific TRAP-seq analysis revealed some unique characteristics of FT-expressing cells, we still do not know their characteristics in a single-cell resolution. As FT is expressed at specific times of the day in relatively small numbers of phloem companion cells, we needed a method to collect the information from these rare FT-expressing cells in a time-dependent manner. To capture FT-expressing cells when FT expression peaks, we deployed single-nuclei RNA-seq (snRNA-seq) combined with fluorescence-activated nuclei sorting (FANS) to isolate GFP-labeled cell-type-specific nuclei at a specific time of the day. Here, we report the unique characteristics of FT-expressing cells at a single-cell level and identify new, growth condition-specific transcriptional repressors of FT.

Results

Bulk nuclei RNA-seq analysis finds profound expression differences in FT-expressing cells between cotyledons and true leaves

To investigate gene expression in the nuclei of specific cell populations, we used Arabidopsis transgenic lines expressing the gene encoding NTF [Nuclear Targeting Fusion protein (16)] under the control of cell type-specific promoters. In addition to the previously generated pFT:NTF line (15), we generated transgenic lines with pSUC2 to capture most phloem companion cells, pCAB2 for mesophyll cells, and p35S as a non-specific control (Fig. 1A). We employed FANS to isolate the respective GFP-positive nuclei (17).

Figure 1.

Although FT is expressed in both cotyledons and true leaves, it was heretofore unknown whether FT-expressing cells in cotyledons and true leaves were equivalent to one another at the whole transcriptome level or showed tissue-specific differences in transcription. The previous histological analyses of the Arabidopsis pFT:GUS reporter line showed that FT expression in true leaves was confined to the distal part of the leaf vasculature, whereas in cotyledons, FT expression was observed in the broader part of the vein (11, 18), suggesting different profiles of FT-expressing cells in cotyledons and true leaves. To exclude possible tissue-specificity as a source of noise in our single-nuclei data, we conducted bulk nuclear RNA-seq of GFP-positive nuclei isolated from either cotyledons or true leaves. To do so, we harvested cotyledons and the first set of two true leaves from two-week-old transgenic plants grown under LD+FR conditions at Zeitgeber time 4 (ZT4), the time of the FT morning expression peak, and isolated GFP-positive nuclei within an hour using a cell sorter (Fig. S1A, C). We successfully collected GFP-positive nuclei from both cotyledons and true leaves for all transgenic lines (Fig. S2) and conducted bulk RNA-seq.

The principal component analysis (PCA) of the resulting expression data indicated separation by tissue and targeted cell population (Fig. 1B). Interestingly, the biggest difference among these datasets was the difference between cotyledons and true leaves, rather than a cell/tissue-type difference. Among different cells/tissues, FT-expressing cells show a distinct gene expression profile compared with other cell/tissue types examined.

Next, to ensure that we had properly enriched for the targeted cell populations, we conducted pairwise comparisons between the p35S:NTF control line and the cell type-specific NTF lines (Data S1–6). As expected, in true leaves of the pSUC2:NTF line, phloem companion cell marker genes FT, SUC2, AHA3, APL, CM3, and AAP4 showed higher expression than in the p35S:NTF line, and the pFT:NTF line showed strong upregulation of FT in both cotyledon and true leaves (Fig. 1C, Fig. S3A–E). True leaves of the pFT:NTF line also showed upregulation of other phloem companion cell marker genes such as AHA3, APL, and AAP4; however, cotyledons of the pFT:NTF line only showed upregulation of AAP4 (Fig. S3A–E), indicating some of known vascular marker genes are more suitable for representing vascular tissues in true leaves.

Similarly, in the pSUC2:NTF line, we observed differences in cell marker gene expression between cotyledons and true leaves. In cotyledons of the pSUC2:NTF line, CM3 and AAP4 were upregulated, but not SUC2, FT, AHA3 and APL. To ensure that FANS properly enriched nuclei from pSUC2:NTF cotyledons, we checked the spatial expression pattern of each gene that was highly expressed in pSUC2:NTF cotyledons [adjusted P-value (padj) < 0.05, fold-change > 2-fold, Table S1]. The majority of the genes upregulated in the cotyledons of our pSUC2:NTF line were found to be expressed in the vasculature. As expected, expression of mesophyll marker gene CAB2 was significantly lower in pSUC2:NTF cotyledon samples (padj = 1.5 x 10⁻⁴), in addition to the mesophyll cell marker RBCS1A in true leaf samples, suggesting that a significant amount of mesophyll cells was successfully removed by FANS (Fig. S3F, G).

Having ensured proper enrichments of the targeted nuclei, we compared global changes in expression among the tested transgenic lines. Consistent with the PCA analysis, the vast majority of differentially expressed genes within a given transgenic line did not overlap between cotyledons and true leaves (Fig. S4, S5). Next, we compared the expression of known FT positive and negative regulators in cotyledons and true leaves of the pFT:NTF, pSUC2:NTF, and pCAB2:NTF lines (Fig. 1D–K) (6).

Among the transcription factors that promote FT expression, APL and VOZ1 were upregulated in true leaves of the pFT:NTF and pSUC2:NTF lines (Fig. 1G). In true leaves of the pSUC2:NTF line, a CO destabilizing factor was significantly upregulated (Fig. 1I), in addition to TFs that repress FT expression (Fig. 1K). In contrast to pSUC2:NTF true leaves, true leaves of the pFT:NTF line did not show significant upregulation for most of these FT repressors. This lack of FT repression appears to be crucial for specifying FT-expressing cells. In contrast, there were far fewer expression differences in cotyledons between the pFT:NTF and pSUC2:NTF lines, indicating that the respective cells in true leaves have diverged more than comparable ones in cotyledons.

Similarly, FT expression and FT transport appeared to be more tightly associated in true leaves than in cotyledons. As reported, the expression of the FT transporter gene FTIP1 is highly associated with FT expression (15). Consistent with this earlier finding, among FT transporter genes, we observed significant upregulation of FTIP1 expression in true leaves of the pSUC2:NTF and the pFT:NTF lines but not in the respective cotyledon samples (Fig. S3H–K).

To further support true leaves as most suitable for investigating FT-expressing cells, we performed Terms enrichment analysis with Metascape (19). In the pFT:NTF line, the genes involved in the “regulation of reproductive process” were upregulated in both cotyledons and true leaves, while the genes related to the term “long-day photoperiodism, flowering” were only upregulated in the true leaf samples of the pFT:NTF line (Fig. S4).

snRNA-seq revealed subpopulation of phloem companion cells that highly express FT

Next, we used true leaves of the pFT:NTF and pSUC2:NTF lines for snRNA-seq using the 10X Genomics Chromium platform (Fig. S1B). We captured a total of 1,173 nuclei for the pFT:NTF line and 3,650 nuclei for the pSUC2:NTF line. In total, we captured 4,823 nuclei and detected 20,732 genes, a median of 149 genes per nuclei (Fig. 2A, B). Gene expression of these nuclei were projected in a two-dimensional UMAP as 11 different clusters (20) (Fig. 2). The nuclei in clusters 8 and 10 showed substantially higher number of transcripts than other clusters (Fig. S6A and B). Because this discrepancy made comparisons difficult, we focused on analyzing gene expression in the other clusters.

Figure 2.

We first asked which nuclei highly expressed FT. These nuclei resided in cluster 7 (Fig. 2D, 2E). SUC2 expression showed a far broader distribution across clusters than FT expression (Fig. 2F). The nuclei in cluster 7 significantly highly expressed 268 genes compared with total population (Data S7). Among these genes, we found FLP1 (Fig. 2G), a gene encoding a flowering-promoting factor acting in parallel with FT (15). Next, we cross-checked these 268 genes with genes identified in the previous TRAP-seq experiment (15). 202 of the 268 genes showed higher expression in FT-expressing cells (pFT:FLAG-GFP-RPL18) than in whole companion cells (pSUC2:FLAG-GFP-RPL18) (Fig. S7A) in the prior study, validating our single nuclei analysis. Further, FKBP12, a gene encoding a CO stabilizing protein was expressed in cluster 7 (Fig. S7B).

To explore the FT-expressing cluster 7, we conducted Terms enrichment analysis using Metascape (19). Genes involved in “Oxidative phosphorylation” were enriched in cluster 7 (Fig. S7C), suggesting that ATP synthesis is particularly active in nuclei belonging to this cluster (Fig. 2H). Genes involved in proton-generating complex I–IV and ATP synthesizing complex V in mitochondrion membrane were also upregulated in cluster 7 (Fig. S8D), suggesting that the entire ATP synthesis pathway is activated in FT-expressing nuclei. Additional terms such as “generation of precursor metabolite and energy” and “purine nucleotide triphosphate metabolic process,” further indicated upregulation of ATP production. Since phloem companion cells actively hydrolyze ATP to generate proton gradients to load sucrose and amino acids (21), it is plausible that FT-expressing phloem companion cells generate a substantial amount of ATP for transport.

We next inquired about the characteristics of other phloem nuclei clusters. Like cluster 7, cluster 4 showed significantly higher SUC2 expression compared with other clusters, suggesting companion cell identity (Fig. 2F, Data S7). Moreover, the genes LTP1, MLP28, and XTH4 were upregulated in cluster 4, consistent with prior studies showing their exclusive expression in vascular tissues including phloem (Fig. S8A–C) (22–24).Terms enrichment analysis found that genes encoding aquaporin were enriched in cluster 4 (Fig. 2I, Fig. 8D, E). Taken together, the nuclei in cluster 4 are likely to be derived from phloem cells that are active in solute transport.

Cluster 5 is mostly composed of nuclei from the pSUC2:NTF line (464 out of 515 nuclei) (Fig. S6B). The nuclei in this cluster showed differentially expressed genes related to defense (Fig. S9A). Specifically, MYC2, a master regulator of response to jasmonic acid (JA), and the JA biosynthetic genes LOX3, LOX4, OPR3, and OPCL1, are expressed in cluster 5 (Fig. S9B). These genes are known to be expressed in vascular tissues, with OPR3 expressed in phloem companion cells and LOX3/4 in phloem (25–27). This subpopulation of phloem cells appears to play pivotal roles in JA-dependent defense responses.

Cluster 6 consisted of both phloem parenchyma and bundle sheath cells (Fig. S10A–F). Subclustering revealed that the nuclei in the upper part of cluster 6 expressed phloem parenchyma marker genes, and the nuclei in the lower part expressed bundle sheath marker genes (28) (Fig. S10C–F). Overall, the genes expressed in this cluster were enriched for genes functioning in sulfur metabolisms (Fig. S10G), including glucosinolate biosynthesis (Fig. S10H). A previous study showed that sulfur metabolic and glucosinolate biosynthetic genes are actively expressed in bundle sheath cells (29).

Lastly, Clusters 1, 2, and 3 showed expression of mesophyll cell marker genes (Fig. S6C–F), suggesting that some mesophyll cells were included during the sorting process. This result might be due to weak induction of the FT and SUC2 promoters in mesophyll cells (15). Indeed, we found both genes to be expressed at low levels in mesophyll protoplasts (Fig. S11). In summary, using snRNA-seq combined with FANS, we revealed the presence of a unique subpopulation of cells with high nuclear FT expression. In these nuclei, genes related to ATP synthesis were enriched. Moreover, we identified other phloem cell clusters enriched for genes in water transport and JA response, which were not previously reported (28).

Next, we subclustered the nuclei in cluster 7, finding three subclusters (Fig. 3). Subcluster 7.2 contained the nuclei with the highest FT expression, in addition to expression of the companion cell marker genes, SUC2 and AHA3. These nuclei contained fewer transcripts of the mesophyll marker genes RBCS1A, CAB3, and CAB2 (Fig. 3C, D), consistent with a prior scRNA-seq study (30). The vascular tissue located at the distal part of true leaves, where FT is expressed highly, is developmentally old, whereas veins emerging from the bottom part of leaves are developmentally young (31). Thus, subcluster 7.3 may be comprised of nuclei from developmentally younger companion cells that have not fully matured yet, whereas those in subcluster 7.2 are older and have gained a stronger companion cell identity.

Figure 3.

Phloem companion cells with high FT expression express other genes encoding small proteins

In LD+FR conditions, FT-expressing cells also express FLP1, which encodes another small protein with systemic effects on flowering and inflorescent stem growth (15). We asked whether the cluster 7 nuclei might express other genes encoding small proteins, which could move through phloem flow. Indeed, the median number of amino acids encoded in the differentially expressed genes of cluster 7 is smaller than those in clusters 4 and 5 (Fig. 4A). To validate whether high FT-expressing vascular cells co-express genes specifically upregulated in cluster 7, which encode small proteins, we selected eight such genes and generated their respective promoter fusions with H2B-tdTomato in the pFT:NTF background. Indeed, the spatial expression patterns of these eight genes and high FT signals overlapped (Fig. 4B).

Figure 4.

Next, we asked whether some of these genes encode new systemic flowering or growth regulators. We selected eight genes, five of which were tested for overlap with FT expression (Fig. 4B) and overexpressed each one of them using the SUC2 promoter. Some of these genes were previously implicated in flowering. BFT is an FT homolog and acts as a floral repressor (32, 33). SENESCENCE-ASSOCIATED AND QQS-RELATED (SAQR) promotes flowering under short-day (SD) conditions (34). RAD-LIKE 4 (RL4) has high homology to the RADIALIS (RAD) gene in Antirrhinum majus, whose ectopic expression in Arabidopsis causes growth defects and late flowering (35). We measured the flowering time of the overexpression lines in independent T1 plants (Fig. 4C). T1 pSUC2:BFT plants grown under LD+FR conditions showed severely delayed flowering compared with pSUC2:GFP control lines (Fig. 4C). This result is in stark contrast to prior results obtained under standard laboratory light conditions (33), suggesting that BFT proteins might be mobile or more functionally active under natural light conditions. None of the other transgenic lines showed noticeable changes in flowering time.

Motif enrichment analysis and transgenic studies identify NIGT1 transcription factors as direct FT repressors

We performed motif enrichment analysis on the promoters of the 268 genes differentially expressed in cluster 7 (36, 37), using randomly selected 3,000 genes as a control set. The most enriched motif was the motif of the NITRATE-INDUCIBLE GARP-TYPE TRANSCRIPTIONAL REPRESSOR 1.2 (also known as HHO2, P = 9.3 x 10⁻⁶) (Fig. 5A; Data S8). Although the NIGT1/HHO transcription factors (TFs) are primarily known as repressors of genes involved in nitrate uptake, constitutive expression of NIGT1.2 significantly decreases FT expression (38, 39). In fact, there are two potential NIGT1 binding sites located within and adjacent to the shadow 1 and 2 domains (S1 and S2) in the FT regulatory region (40) (Fig. S12). These domains are highly conserved among Brassicaceae species and important for FT transcriptional regulation. Moreover, the NIGT1 binding motifs reside in close proximity to the CO binding motif (TGTGNNATG, CO-responsive element, CORE) (40, 41), suggesting that NIGT1 binding could affect CO binding.

Figure 5.

We performed a yeast one-hybrid (Y1H) screen of 1,957 TFs against four tandem repeats of a short FT promoter sequence containing the S1 and S2 domains and the two NIGT1 binding sites (42, 43). We found that NIGT1.1/HHO3, NIGT1.2/HHO2, NIGT1.3/HHO1, NIGT1.4/HRS1, and HHO5 specifically bind to this sequence within the FT regulatory region (Data S9). Next, we overexpressed five different NIGT1/HHO genes using the SUC2 promoter. The overexpression of NIGT1.2 and NIGT1.4 delayed flowering, while the overexpression of NIGT1.1, NIGT1.3, and HHO5 did not (data not shown). Therefore, we generated T3 lines of pSUC2:NIGT1.2 and pSUC2:NIGT1.4 and measured their flowering time. Under LD+FR conditions, these transgenic plants flowered significantly later than wild-type plants (Fig. 5A). To test which genes were altered by the overexpression of NIGT1.2 and NIGT1.4, we conducted RNA-seq analysis of the respective transgenic plants (Fig. 5B). We found that genes specifically expressed in cluster 7 such as FT, BFT, and SAQR were downregulated in both the pSUC2:NIGT1.2 and pSUC2:NIGT1.4 lines. FLP1 expression was decreased only in pSUC2:NIGT1.2 lines, which might explain why pSUC2:NIGT1.2 plants exhibited a more severe late flowering phenotype than pSUC2:NIGT1.4 plants. Similar results were obtained with qRT-PCR analysis (Fig. S13). Consistent with our snRNA-seq data (Fig. S14), previous studies reported that NIGT1.2 is expressed broadly in shoots while NIGT1.4 is expressed only in roots (38), suggesting that NIGT1.2 is the major player in FT regulation.

Ample nitrogen prolongs the vegetative growth stage and delays flowering (44); however, the detailed mechanisms of this delay remain largely unknown (44). Since NIGT1 genes are induced under high nitrogen conditions, we tested the quadruple mutant of the NIGT1 genes (nigtQ mutant) for FT and BFT expression with (+N) and without high nitrogen (–N) (38). The nigtQ plants showed enhanced FT and BFT expression compared to WT plants under +N but not –N conditions, suggesting that the NIGT1 TFs act as novel nitrate-dependent repressors of flowering (Fig. 5C).

Discussion

Plants utilize seasonal information to determine the onset of flowering. One of the key mechanisms of seasonal flowering is the transcriptional regulation of the florigen FT in leaves. Our previous study revealed that FT-expressing cells also produce FLP1, another small protein that may systemically promote both flowering and elongation of leaves, hypocotyls, and inflorescence stems (15). Therefore, a more detailed characterization of FT-expressing cells might identify additional components that are essential for the integration of the many environmental cues in natural environments into the precise timing of flowering and related developmental changes. Here, we applied bulk nuclei and single-nuclei RNA-seq and transgenic analysis to examine FT-expressing cells at high resolution.

Differences in the spatial expression patterns of FT between cotyledons and true leaves are likely driven by negative regulators

In young Arabidopsis plants grown in LD, true leaves show FT expression in the distal and marginal parts of the leaf vasculature, while cotyledons express FT across entire veins (11, 18). We used bulk RNA-seq to examine the FT-expressing cells in cotyledons versus those in true leaves, using FANS-enriched nuclei. Tissue-specific differences (cotyledons vs. true leaves) in gene expression were more significant than those observed for the enriched cell populations, an unexpected result as cotyledons and true leaves are not typically treated as separate organs.

The true leaves but not the cotyledons of the pSUC2:NTF and pFT:NTF lines showed differences in the expression of FT negative regulators. This result suggests that FT expression is strongly repressed in most of the true leaf companion cells, contributing to the spatial expression pattern of FT observed in true leaves. In this study, we focused on the FT expression peak in the morning (ZT4) under LD+FR conditions; however, in the future, it would be interesting to investigate the expression patterns of FT regulators at the FT evening peak (ZT16), as our results suggest that there are two different mechanisms regulating each FT peak in natural long days (13, 14).

A subpopulation of phloem companion cells highly expresses FT and other genes encoding small proteins

Our snRNA-seq identified a cluster with nuclei derived from cells with high FT expression. The cluster 7 nuclei are derived from cells that appear to be metabolically active and produce ATP, which may be used to upload sugars, solutes, and small proteins, including FT, into the phloem sieve elements. Based on subclustering analysis, FT expression levels were positively correlated with companion cell marker gene expression and negatively correlated with mesophyll cell marker gene expression. The observed shift from mesophyll to companion cell identity may represent the trajectory of companion cell development (30).

In clusters 4 and 5, we found nuclei isolated from phloem cells with high expression of aquaporin and JA-related genes, respectively. These cell populations were not identified in a previous protoplast-based scRNA-seq analysis targeting phloem (28). We speculate that these cells are highly embedded in the tissue and therefore practically difficult to isolate. Orthologous studies show that JA biosynthetic genes such as LOX3/4 and OPR3 are highly expressed in phloem including companion cells (25, 27), consistent with our results.

Aquaporin genes such as PIP2;1 and PIP2;6 are thought to be highly expressed in leaf xylem parenchyma and bundle sheath tissues (45). However, cluster 6, which contains nuclei isolated from bundle sheath and phloem parenchyma cells, showed fewer expressed aquaporin genes than the cluster 4 nuclei, which also expressed SUC2. This result might indicate that the cells active in water transport exist not only in bundle sheath or xylem parenchyma cells but also in a subpopulation of phloem cells. In summary, SUC2/FT-expressing vasculature cells consist of cell subpopulations with different functions within true leaves.

BFT may fine-tune the balance between flowering and growth as a systemic anti-florigen

The FT-expressing nuclei of cluster 7 preferentially expressed genes encoding small proteins including FLP1, BFT, and SAQR (32–34). Although a previous study reported that BFT driven by the SUC2 promoter does not alter flowering time (33), our result shows that overexpression of this gene delays flowering under LD+FR conditions (Fig. 4C). This discrepancy is likely due to our use of growth conditions with the adjusted R/FR ratio that mimics natural light conditions, as the majority of our transgenic lines showed a similar late flowering phenotype.

Why do Arabidopsis plants produce florigen FT and anti-florigen BFT in the same cells? There is precedent for the co-expression of florigen and anti-florigen in leaves in other plants. In Chrysanthemums, a weak florigen CsFTL1 and a strong anti-florigen CsAFT are both expressed in leaves in long days. Short-day conditions repress CsAFT but induce the expression of a strong florigen CsFTL3 to initiate flowering (46, 47). The balance in florigen and anti-florigen expression is also important for wild tomatoes to control flowering time, and this regulation was lost in domesticated tomatoes, resulting in day-neutral flowering behaviors (48). In Arabidopsis, the FT-homolog floral repressor TERMINAL FLOWER 1 (TFL1) is expressed at the shoot apical meristem and competes with FT for physical interaction with FD (49). In addition to preventing precocious flowering, the presence of TFL1 is crucial for the balance between flowering initiation and stem growth because FT terminates inflorescence stem growth (50). Like FT and TFL1, BFT directly binds to FD (33), suggesting that BFT’s mode of action is similar to that of TFL1. However, unlike for TFL1, loss of BFT does not affect flowering time under standard growth conditions (32, 33). The lack of a visible flowering phenotype in the bft knockout mutant might be due to the substantially lower expression of BFT compared to FT under standard growth conditions (33). Consistent with this interpretation, the bft knockout mutant is significantly delayed in flowering under high salinity conditions where BFT gene expression is strongly induced (33). Thus, BFT appears to act as an anti-florigen under specific growth conditions, possibly including LD+FR conditions. Given the complexity of natural environments, the existence of multiple specialized anti-florigens might facilitate signal integration and reduce noise in the onset of flowering and stem growth.

NIGT1 TFs contribute to the nutrient-dependency of flowering time

To identify potential novel transcriptional regulators of FT, we identified enriched motifs in the promoters of the 268 genes that were differentially expressed in cluster 7. This analysis indicated that NIGT1 TFs may affect the expression of FT and other genes co-expressed with FT. Using Y1H screening, we found that all NIGT1s and HHO5 can bind to the enhancer sequences derived from the FT promoter. NIGT1s are involved in nitrogen absorption as negative regulators (38, 39). It is well-known that nitrogen availability affects the onset of flowering; however, most of the mechanistic underpinnings remain unknown. A recent study showed that ample nitrogen availability causes phosphorylation of FLOWERING BHLH 4 (FBH4), a direct positive regulator of CO, which results in attenuation of FBH4 transcriptional activity (44). This mechanism acts upstream of FT, whereas the NIGT1 TFs likely act as direct repressors of FT. Multiple nitrogen-dependent mechanisms acting on FT regulation might ensure the proper balance between nutrient availability and resource allocation toward developmental transitions.

Materials and Methods

Molecular cloning and plant materials

All Arabidopsis thaliana transgenic plants and mutants are Col-0 backgrounds. ACT2:BirA plant (16) was used as the Arabidopsis genetic background to generate NTF expressing lines. The NTF cDNA was controlled by the 35S promoter or the tissue-specific promoters: CAB2 (AT1G29920), SUC2 (AT1G22710), and FT (AT1G65480). The NTF sequences were amplified using primers (5’-CACCATGGATCATTCAGCGAAAACCACACAG-3’ and 5’-TCAAGATCCACCAGTATCCTCATGC-3’) and cloned into the pENTR/D-TOPO vector (Invitrogen). The CAB2 promoter region (324 bp) was amplified by the primers (5’-CACCATATTAATGTTTCGATCATCAGAATC-3’ and 5’-TTCGATAGTGTTGGATTATATAGGG-3’) and cloned into the pENTR 5’-TOPO (Invitrogen) (51). The SUC2 promoter region (2,302 bp) was amplified by primers (5’-GGTGCATAATGATGGAACAAAGCAC-3’ and 5’-ATTTGACAAACCAAGAAAGTAAG-3’) and cloned into the pENTR 5’-TOPO. The CAB2 and SUC2 promoter sequences were fused with NTF in the binary GATEWAY vector R4pGWB501 through LR clonase reaction (52). The CAB2:NTF and SUC2:NTF constructs were transformed into wild-type Col-0. The pFT:NTF line was described in the previous study (15).

The H2B-tdTomato constructs containing promoter regions of cluster 7 highly expressed genes were made by swapping the heat-shock promoter (pHS) of pPZP211 HS:H2B-tdTomato (15). Promoter sequences were amplified by the forward primer containing SbfI site and reverse primer containing SalI site, and inserted into these restriction enzyme sites. Following primer sequences were used to amplify 2,396 bp upstream of BFT (AT5G62040), 5’-CCTGCAGGGACAGAGTAAATTCAACCACAGCAGGT-3’ and 5’-GTCGACTTTTCTTTGCTCCAATGTGTTTGCGTTTG-3’; 2,521 bp upstream of AT1G24575, 5’-CCTGCAGGCTCTCAGATCACCGTAAGGGCATAATTATATTTAGGTTCAC-3’ and 5’-GTCGACGTGATGAGATTTGTGACTGGAGGAGTTTCCAAGTACCATTCTT-3’; 2,488 bp upstream of AT1G67865, 5’-CCTGCAGGACTTCACATTCTTGGATTCCGTTTGTAATAACTAATGTTTT-3’ and 5’-GTCGACCCCTCCGGCAACCCCAATAATAAGCTTATCAAGCATTTTTCTT-3’; 1,939 bp upstream of ROXY10 (AT5G18600), 5’-CCTGCAGGGCAATGGACCGTACGTCTAGGTCACGCATCTTATCCGACAT-3’ and 5’-GTCGACCACCGGTCTCTCCATCACCATCTTCGTTATCATATCCATTGCT-3’; 2,226 bp upstream of AT2G26695, 5’-CCTGCAGGAATGTAATGTATAATGTGTTCATAAACAGCACCAACTACCC-3’ and 5’-GTCGACTGCGTGCTGACACGCACCACATAGCCAATCTCCTCCGGTCCAG-3’; 1,911 bp upstream of PARCL (AT1G64370), 5’-CCTGCAGGGCCCATCTAATTCCCATTTTAGATGCATGAGTTCAACGCTA-3’ and 5’-GTCGACGCCTTGAGCCACCTCGTAGTAGTCTTTCTCACGGTTTTCGTAG-3’; 2,719 bp upstream of ROXY7 (AT2G30540), 5’-CCTGCAGGATCACCGGTAAGTGACAAGAGAATTGA-3’ and 5’-GTCGACGGTTTCTTGAAGGAGGTCTCGATCAATCT-3’; 2,574 bp upstream of CYSTM12 (AT5G04080), 5’-CCTGCAGGGAGAATTTGAAGGAGGCTTTGCGTTTTATCTGCTCATCTAA-3’ and 5’-GTCGACTTGTGGAGGATTTTGATCTCTCATGTCCTGCATCTTCTCAAAA-3’. These constructs were transformed into the pFT:NTF plants.

To overexpress cluster 7 highly expressed genes, we amplified cDNA using the following primer sets. BFT, 5’-CACCATGTCAAGAGAAATAGAGCCAC-3’ and 5’-AGTTAATAAGAAGGACGTCGTCG-3’; AT1G24575, 5’-CACCATGGTACTTGGAAACTCCTCCAGTCAC-3’ and 5’-GACTAGGCGCTCTTAGTCATCCAC-3’; AT1G67865, 5’-CACCATGCTTGATAAGCTTATTATTGGGGTTGCC-3’ and 5’-CTTTACTCCCTGTCTTTCTGGCG-3’; ROXY10, 5’-CACCATGGATATGATAACGAAGATGGTGATGGAG-3’ and 5’-AGTCAAACCCACAATGCACCAG-3’; AT2G26695, 5’-CACCATGAGCTGGACCGGAGGAG-3’ and 5’-ATTTAGACGCCACCATAATCTCTTG-3’, SAQR (AT1G64360), 5’-CACCATGTCGTTTAGAAAAGTAGAGAAGAAACC-3’ and 5’-GATTAGTAATTAGGGAAGTGTTTGCGGC-3’; RL4 (AT2G18328), 5’-CACCATGGCTTCTAGTTCAATGAGCACC-3’ and 5’-CTTCAATTAGTGTTACGGTACCTAGG-3’; AT1G54575, 5’-CACCATGGTGGATCATCATCTCAAAGC-3’ and 5’-AATTAATTATTCTTTTGTGGCTTGG-3’. Amplified cDNA was cloned into pENTR/D-topo, followed by GATEWAY cloning using pH7SUC2, a binary vector carrying 0.9 kb SUC2 promoter (15). These vectors were transformed into wild-type plants possessing pFT:GUS reporter gene (11).

For gene expression measurements and confocal imaging, surface sterilized seeds were sawn on the 1x Linsmaier and Skoog (LS) media plates without sucrose containing 0.8% (w/v) agar and stratified for more than 2 days at 4 °C before being transferred to the incubator. Plants were grown under LD+FR conditions (100 µmol photons m⁻² s⁻¹, red/far-red ratio = 1.0) for two weeks as described previously (6). For gene expression analysis under –N conditions, plants were grown on normal 1x Murashige and Skoog (MS) medium plates with full nitrogen (+N) for 10 days and transplanted to +N and –N medium plates and grown for 4 additional days. The ion equilibrium of the medium between +N and –N was ensured by replacing KNO3 (18.79 mM) and NH4NO3 (20.61 mM) by KCl (18.79 mM) and NaCl (20.61 mM). For flowering time measurements using T1 transgenic lines (Fig. 4B), T1 plants were grown on hygromycin selection plates for 10 days and transferred to Sunshine Mix 4 soils (Sun Gro Horticulture). Soils were supplemented with a slow-release fertilizer (Osmocote 14-14-14, Scotts Miracle-Gro) and a pesticide (Bonide, Systemic Granules) and filled in standard flats with inserts (STF-1020-OPEN and STI-0804, T.O. Plastics). For the same experiment using pSUC2:NIGT1 T3 plants (Fig. 5B), surface sterilized seeds were directly sawn on soils and kept in 4 °C for at least 2 days. Plants on soils were kept grown under LD+FR conditions as described previously (6). Tissue clearing and confocal imaging were conducted as previously described (6).

SMART-seq2 library preparation and analysis

For bulked nuclei RNA-seq, approximately 15,000 nuclei were collected from cotyledons and true leaves of p35S:NTF, pCAB2:NTF, and SUC2:NTF lines, while 10,000 (cotyledons) and 3,000 (true leaves) nuclei were collected from pFT:NTF due to fewer population of GFP-positive nuclei. Sorted nuclei were directly collected into Buffer RTL (Qiagen), and RNA was extracted according to the manufacture protocol of RNeasy Micro Kit (Qiagen). Three independent biological replicates were produced for cotyledon and true leaf of all transgenic lines.

After RNA integrity was confirmed using High Sensitivity RNA Screen Tape (Agilent), SMART-seq2 libraries were generated according to the manufacturer’s protocol (Clontech). Sequencing was performed on the Illumina NovaSeq 6000 platform through a private sequencing service (Novogene). Since reverse reads were undesirably trimmed due to the presence of in-line indexes in SMART-seq libraries, we used only forward reads of paired-end reads. Approximately 10.8 to 34.6 million forward reads (average, 20.2 million reads) were produced from each sample. By STAR software (version 2.7.6a) (53), reads were mapped to Arabidopsis genome sequences from The Arabidopsis Information Resource (TAIR, version 10) (http://www.arabidopsis.org/). DEGs were selected based on the adjusted P-value calculated using DEseq2 in the R environment.

Single-nucleus RNA-seq library preparation

For snRNA-seq experiments, sorted nuclei from true leaves were collected in the mixture of 100 µL 1x PBS and 10 µL SUPERase-In RNase inhibitor in a 15 mL corning tube. Prior to nuclei collection, the corning tube was coated by rotating with 10 mL 1x PBS containing 1% BSA at room temperature to prevent static electricity in the wall. After more than 10,000 nuclei were collected in the corning tube, 10 µL of 1 mg/mL DAPI solution and 1/100 volume of 20 mg/mL glycogen (Thermo Scientific) were added and centrifugated in 1,000xg for 15 min at 4 °C. Nuclei numbers were determined using a hemocytometer after resuspension with a small volume of 1x PBS. The total number of nuclei varied depending on the plant line and trial but was no more than 11,000.

snRNA-seq was performed using the 10x Single-cell RNA-Seq platform, the Chromium Single Cell Gene Expression Solution (10x Genomics). Two biological replicates of SUC2:NTF and one replicate of pFT:NTF were produced for a total of three samples.

Estimating gene expression in individual nucleus

Sequencing of snRNA-seq reads was performed on the Illumina Nextseq 550 platform, followed by the mapping to the TAIR10 Arabidopsis genome using Cellranger version 3.0.1 software.

The Seurat R package (version 4.0.5) (54, 55) was used for the dimensional reduction of our SnRNA-seq data. To remove potential doublets and background, we first filtered out nuclei with less than 100 and more than 5,500 detected genes, and more than 20,000 UMI reads. For UMAP data visualization and cell clustering, all biological replicates were combined, and twenty principal components were compressed using a resolution value of 0.5.

Significantly highly expressed genes in each cluster compared with entire nuclei population were identified using FindMarkers in Seurat with default parameters.

To show the enrichments of specific sets of genes, average expression of ATP biosynthesis, JA responsive, aquaporin, bundle sheath and phloem parenchyma marker genes were visualized using AddModuleScore in Seurat. For bundle sheath and phloem parenchyma markers, we leveraged the top 50 most highly expressed genes in these cell types in previous protoplast-based ScRNA-seq data (28).

Protein amino acid length

Amino acid length of proteins encoded by genes highly expressed in clusters 4, 5, and 7 were obtained from the Bulk Data Retrieval tool in TAIR database (https://www.arabidopsis.org/tools/bulk/sequences/index.jsp).

Promoter cis-enrichment analysis

To identify highly enriched cis-elements in cluster 7 highly expressed genes, promoter sequences of 268 genes highly expressed in cluster 7 and randomly selected 3,000 genes by R coding were extracted using the Eukaryotic Promoter Database (36, 37). Obtained promoter sequences were next submitted to Simple Enrichment Analysis (SEA) (56) in MEME Suite server (version 5.4.1) to elucidate what cis-elements are enriched in cluster 7 highly expressed genes through a comparison with random 3,000 genes.

Yeast one-hybrid screening

Four tandem repeats of S1/S2 elements of FT promoter that include a CO-responsive (CORE) element (40, 41) were generated through restriction enzyme-mediated ligation. Forward and reverse oligonucleotides containing S1/S2 element with HindIII (CORE-F1: 5’-AGCTTACTGTGTGATGTACGTAGAATCAGTTTTAGATTCTAGTA CTGTGTGATGTACGTAGAATCAGTTTTAGATTCTAG-3’), EcoRI (CORE-R1: 5’-AATTCCTAGAATCTAAAACTGATTCTACGTACATCACACAGTACTAGAATCTAAAACTGATTCT ACGTACATCACACAGTA-3’), SacI (CORE-F2: 5’-CACTGTGTGATGTACGTAGAATCAGTTTTAGATTCTAGTACTGTGTGATGTAC GTAGAATCAGTTTTAGATTCTAGGGTAC-3’), and KpnI (CORE-R2: 5’-CCTAGAATCTAAAACTGATTCTACGTACATCACACAGTACTAGAATCTAAAACTGATTCTACG TACATCACACAGTGAGCT-3’) recognition sequences were chemically synthesized from Genewiz (https://www.genewiz.com/). The oligonucleotides complementary to each other were denatured for 10 min at 95 ⁰C, then annealed for 30 min at RT. The resultant product was inserted into pENTR/D-TOPO vector harboring MCS sites (HindIII, EcoRI, SacI and KpnI) through restriction enzyme-mediated ligation. Four tandem S1/S2 elements in pENTR/D-TOPO were then inserted into pY1-gLUC59-GW vector (42) using LR clonase II (Invitrogen). The resultant pY1-gLUC59-GW-2XCORE plasmids were used for Y1H screening analysis

qRT-PCR analysis using total RNA

Total RNA extraction, cDNA synthesis and qRT-PCR were performed as previously described (6). ISOPENTENYL PYROPHOSPHATE / DIMETHYLALLYL PYROPHOSPHATE ISOMERASE (IPP2) and PROTEIN PHOSPHATASE 2A SUBUNIT A3 (PP2AA3) were used as reference genes. For statistical tests, relative expression levels were log₂-transformed to meet the requirements for homogeneity of variance. qPCR primers used in this study are listed in Table S2.

RNA-seq analysis using total RNA

Two-week-old seedlings of pSUC2:NIGT1.2 and NIGT1.4 grown on 1xLS media plates under LD+FR conditions were harvested at ZT4. RNA extraction and RNA-seq library preparation were conducted as described previously (15, 57). As a reference, our previous gene expression data of wild-type plants grown at the exact same conditions was compared with pSUC2:NIGT1 lines (15).

Data availability

Bulk RNA-seq from sorted nuclei can be found at the NCBI short read archive bio project PRJNA1098062. Single-nuclei RNA-seq can be found at NCBI GEO under GSE273032. The RNA-seq data of pSUC2:NIGT1.2 and pSUC2:NIGT1.4 plants have been deposited in the DNA Data Bank of Japan (DDBJ) Sequence Read Archive under PRJDB17784.

Acknowledgements

We thank N.M. Belliveau and D.W Galbraith for technical advice for FACS, Y. Mizuta for providing research materials, M. Ashikari, H. Tsuji, K. Nagai, and M. Mizutani for scientific discussion. We also thank T. Niwa, A. Furuta, S. Ishikawa and T. Takagi for technical assistance. This research was supported by grants from the National Institutes of Health grant (R01GM079712 to J.T.C., C.Q., and T.I.; NIGMS MIRA grant no. 1R35GM139532 to C.Q.), from the National Science Foundation (PlantSynBio grant no. 2240888 to C.Q), Japan MEXT KAKENHI (JP20H05910 and JP22H04978 to T.I.), JSPS KAKENHI (24K18139 to H.T.) and National Science Foundation grant (NSF-IOS #1755452 to J.L.P.-P.).

Author Contributions

H.T. and T.I conceptualized the study.

Methodology: H.T. and C.M.A.

Investigation: H.T., S.I, J.S.S, A.K., A.K.H., N.L. and T.S.

Formal analysis: H.T. T.S. J.L.P.-P. and J.T.C.

Data curation: H.T. T.S. and J.T.C.

Validation: H.T., A.K.H., and T.I.

Visualization: H.T.

Funding acquisition: H.T., J.L.P.-P., J.T.C., C.Q. and T.I.

Resources: S.I., D.K., T.K. and J.L.P.-P.

Project administration: H.T., C.Q. and T.I.

Supervision: Y.S., Y.T., J.L.P.-P., C.Q. and T.I.

H.T., S.I, J.S., A.K.H, J.T.C., and C.Q. and T.I. wrote the manuscript. N.L. reviewed and commented on the manuscript.

Competing Interest Statement

The authors declare no competing interest.