NIGT1 family proteins exhibit dual mode DNA recognition to regulate nutrient response-associated genes in Arabidopsis

Fine-tuning of nutrient uptake and response is indispensable for maintenance of nutrient homeostasis in plants, but the details of underlying mechanisms remain to be elucidated. NITRATE-INDUCIBLE GARP-TYPE TRANSCRIPTIONAL REPRESSOR 1 (NIGT1) family proteins are plant-specific transcriptional repressors that function as an important hub in the nutrient signaling network associated with the acquisition and use of nitrogen and phosphorus. Here, by yeast two-hybrid assays, bimolecular fluorescence complementation assays, and biochemical analysis with recombinant proteins, we show that Arabidopsis NIGT1 family proteins form a dimer via the interaction mediated by a coiled-coil domain (CCD) in their N-terminal regions. Electrophoretic mobility shift assays defined that the NIGT1 dimer binds to two different motifs, 5'-GAATATTC-3' and 5'-GATTC-N38-GAATC-3', in target gene promoters. Unlike the dimer of wild-type NIGT1 family proteins, a mutant variant that could not dimerize due to amino acid substitutions within the CCD had lower specificity and affinity to DNA, thereby losing the ability to precisely regulate the expression of target genes. Thus, expressing the wild-type and mutant NIGT1 proteins in the nigt1 quadruple mutant differently modified NIGT1-regulated gene expression and responses towards nitrate and phosphate. These results suggest that the CCD-mediated dimerization confers dual mode DNA recognition to NIGT1 family proteins, which is necessary to make proper controls of their target genes and nutrient responses. Intriguingly, two 5'-GATTC-3' sequences are present in face-to-face orientation within the 5'-GATTC-N38-GAATC-3' sequence or its complementary one, while two 5'-ATTC-3' sequences are present in back-to-back orientation within the 5'-GAATATTC-3' or its complementary one. This finding suggests a unique mode of DNA binding by NIGT1 family proteins and may provide a hint as to why target sequences for some transcription factors cannot be clearly determined.


Introduction
Plants are continuously exposed to fluctuations in soil nutrient levels because of changes in climatic conditions, microbial activities, and water flow [1][2][3]. To maintain nutrient homeostasis and ensure successful reproduction, plants employ a wide variety of responses that modulate nutrient uptake and metabolism. Response triggered by the shortage of phosphorus (P), one of the major macronutrients needed for plant growth, is one such response [4,5]. Under P-deficient conditions, plants increase the uptake of inorganic phosphate (Pi), the plant-accessible form of P, to maintain internal P concentration [5,6]. In addition, many plant species accumulate foliar anthocyanins under Pi-deficient conditions to effectively dissipate excess light energy and mitigate oxidative stress [5,6]. These responses, collectively termed as Pi starvation responses (PSRs), are induced by a Pi-dependent transcriptional regulation. Similarly, shortage of nitrogen (N), another macronutrient required in large amounts, also triggers a nutrient response. N shortage causes chlorosis and decreases photosynthesis and plant growth [7,8]. Most of the terrestrial plants uptake N in the form of nitrate [9], and respond to nitrate supply as well as long-term depletion of N via transcriptional reprogramming [10,11].
In Arabidopsis thaliana and rice (Oryza sativa L.), some of the GOLDEN2/ARR-B/Psr1 (GARP)-type transcription factors (TFs), such as PHOSPHATE STARVATION RESPONSE 1 (PHR1) and its homologs, play a major role in the initiation of PSR [5,6,[12][13][14]. Although PHR1 gene expression is hardly responsive to fluctuations in Pi availability [6,14], Pi-dependent interaction of PHR1 with SYG1/Pho81/XPR1 (SPX) domain-containing proteins (SPX proteins) modulates PHR1 activity. Under Pi-sufficient conditions, SPX proteins bind to PHR1, thus maintaining PHR1 in an inactive state [15][16][17]. However, upon Pi deprivation, the interaction between SPX proteins and PHR1 is weakened, and PHR1 is released from the PHR1-SPX complex; the released PHR1 then binds to the cis-element GNATATNC in target gene promoters, thus inducing their expression and initiating PSR [15,[17][18][19]. On the other hand, the response to nitrate supply is orchestrated by NIN-LIKE PROTEIN (NLP) transcriptional activators in Arabidopsis and other plant species [11,[20][21][22]. The NLP family proteins induce the expression of a range of N uptake or metabolism-related genes, such as NRT2.1, which encodes a high affinity nitrate transporter [22][23][24][25]. The NLP family proteins also regulate the expression of genes encoding other TFs, such as NITRATE-INDUCIBLE GARP-TYPE TRANSCRIPTIONAL REPRESSOR 1 (NIGT1) family proteins, to generate a complex family proteins and HRS1 HOMOLOG (HHO) family proteins, which are closely related to NIGT1 family proteins but encoded by nitrate-non-inducible genes. The results revealed a conserved amino acid sequence representing a coiled-coil motif in the N-terminal region of NIGT1 and HHO family proteins (S1 Fig). Amino acid sequence alignment of the predicted CCDs of 22 NIGT1 family proteins from 13 plant species reported previously [26] revealed that the domain includes heptad repeats of branched-chain amino acids (leucine [Leu], isoleucine [Ile], and valine [Val]) spanning 22 amino acid residues (Fig 1A). This structural feature resembles that of the Leu zipper domain [37], which is frequently involved in the dimerization of DNA-binding proteins. We, therefore, examined the protein-protein interactions by yeast two-hybrid (Y2H) assays using four NIGT1 family proteins (NIGT1. 1-1.4) and three HHO family proteins (HHO4-6) from Arabidopsis. NIGT1.1 interacted with all four NIGT1 family proteins as well as with HHO4 and HHO6, but not with HHO5 ( Fig 1B). Examination of all pairwise combinations of proteins revealed that the four NIGT1 family proteins and HHO4 interacted with each other; however, their interaction with HHO5 and HHO6 was observed only in certain combinations, suggesting that HHO5 and HHO6 perform distinct functions compared with NIGT1 family proteins and HHO4 (Fig 1C; S2 Fig). In this assay, truncated PHR1 carrying the GARP domain and CCD (208-362 amino acids [aa]) was used as a negative control, because full-length PHR1 forms a homo dimer via a CCD in its C-terminal region [14] and activates transcription in yeast [17]. PHR1 interacted with itself but did not interact with any of the NIGT1 and HHO family proteins, confirming the specificity of the NIGT1-NIGT1 interactions (Fig 1B and 1C; S2 Fig).
Next, to investigate whether the predicted CCD is necessary for NIGT1-NIGT1 interaction, we performed Y2H assays using deletion variants of NIGT1.1 (Fig 1D). The results indicated that the N-terminal region (1-51 aa) is required and sufficient for NIGT1-NIGT1 interaction, whereas other regions are not involved in this interaction (Fig 1E). Among the branchedchain amino acids comprising the heptad repeats in the CCD, two Leu residues (amino acid positions 25 and 39 in NIGT1.1) and an Ile residue (amino acid position 32 in NIGT1.1) were conserved among all examined NIGT1 proteins ( Fig 1A). Positions of these amino acid residues correspond to those of hydrophobic amino acid residues that constitute the hydrophobic surface of the Leu zipper domain [37][38][39]. Thus, we examined whether the wild-type NIGT1.1 protein (NIGT1.1 WT ) interacts with mutant NIGT1.1 proteins in which the 25 th and/or 39 th Leu residue was substituted by the non-branched amino acid alanine (Ala) (NIGT1.1 L25A , NIGT1.1 L39A , and NIGT1.1 L25A/L39A ). The Y2H assays showed a drastic reduction in the NIGT1.1-NIGT1.1 interaction in these mutant NIGT1.1 proteins (Fig 1F). A similar impairment in protein-protein interaction was observed when NIGT1.3 was used as an interacting partner for the mutant NIGT1.1 proteins (Fig 1F). These results suggest that Leu residues in the N-terminal CCD are essential for homomeric and heteromeric interactions between NIGT1 family proteins.

PLOS GENETICS
The NIGT1 dimer mediates nutrient response in Arabidopsis NIGT1.1 WT -nGFP or NIGT1.1 WT -cGFP was co-expressed with cGFP or nGFP fusion of an unrelated nuclear-localizing protein GRC3 [40] (Fig 2A). Furthermore, although nuclear localization of NIGT1.1 WT and NIGT1.1 L25A/L39A was confirmed by fusing them to full-length GFP (Fig 2B), NIGT1.1 L25A/L39A failed to interact with NIGT1.1 WT in leaf cells (Fig 2C). These results demonstrate the interaction among NIGT1 family proteins via the N-terminal CCD in plant cell nuclei, and confirm that the two Leu residues in the CCD are essential for this interaction.

NIGT1.1 forms a dimer via the CCD in solution
We hypothesized that the loss of protein-protein interaction ability of NIGT1 family proteins would interfere with the NIGT1-NIGT1 complex formation, thus affecting their dual mode DNA recognition ability. To test this hypothesis, we produced histidine (His)-and thioredoxin (Trx)-tagged NIGT1.1 WT and NIGT1.1 L25A/L39A proteins in Escherichia coli using pET32 plasmid, and purified these recombinant proteins ( Fig 3A). It is important to note that the bacterial Trx exists as a monomer [41]. Size exclusion chromatography coupled to multi-angle scattering (SEC-MALS) analysis was performed to determine the absolute molecular masses of recombinant NIGT1.1 WT and NIGT1.1 L25A/L39A proteins in solution ( Fig 3B). In this analysis, NIGT1.1 WT protein produced a single peak corresponding to a molecular mass of 123 kDa, which is approximately twice the expected molecular mass of recombinant NIGT1.1 protein (57.4 kDa). On the other hand, NIGT1.1 L25A/L39A protein was mostly eluted at a peak corresponding to a molecular mass of 58.8 kDa. These results revealed that NIGT1.1 WT exclusively exists as a dimer, while most of NIGT1.1 L25A/L39A exists as a monomer.

Disruption of the CCD affects dual mode DNA recognition by NIGT1.1
To test whether the inability of NIGT1 proteins to form dimers affects their dual mode DNA recognition and binding ability, we performed electrophoretic mobility shift assays (EMSAs) using recombinant NIGT1.1 WT and NIGT1.1 L25A/L39A proteins. Previous studies showed that NIGT1 family proteins bind to specific sequences in SPX1 and NRT2.1 gene promoters, and

PLOS GENETICS
The NIGT1 dimer mediates nutrient response in Arabidopsis also revealed the physiological significance of this binding [24,30]. Based on the information from these studies, we used three DNA probes (P1-P3) corresponding to the authentic NIGT1 binding regions that were experimentally identified in SPX1 and NRT2.1 gene promoters: P1 probe, amplified from the SPX1 promoter, contained two copies of the non-palindromic motif (GAATC or its reverse complement, GATTC) and one copy of the palindromic motif (GAA-TATTC); P2 probe, also amplified from the SPX1 promoter, contained only one copy of the palindromic motif; and P3 probe, amplified from the NRT2.1 promoter, contained two copies of the non-palindromic motif ( Fig 4A). Consistent with previous results, the binding of NIGT1.1 WT to these probes was inhibited by the presence of excessive amounts of non-labeled probes as competitors ( Fig 4B). Protein-DNA complexes comprising NIGT1.1 L25A/L39A migrated faster in the gel compared with those comprising the NIGT1.1 WT protein, as expected, because of their lower molecular weight. Furthermore, the amount of NIGT1.1 L25A/ L39A -DNA complexes was less than that of the NIGT1.1 WT -DNA complex, suggesting the compromised affinity of the NIGT1.1 L25A/L39A toward any of the three DNA probes. These results suggest that dimerization is required for the dual mode, high affinity DNA-binding ability of NIGT1 family proteins.
Unique mode of DNA recognition by NIGT1.1 NIGT1.1 WT , which exclusively existed as a dimer, bound to the P3 probe containing only nonpalindromic motifs more effectively, compared with NIGT1.1 L25A/L39A that existed as a . PP2A served as a negative control. Two-tailed Student's t-test was conducted to determine significant differences in the fold enrichment between PP2A and each DNA region, and the significance is indicated as follows; � , P < 0.05; �� , P < 0.001. For region R3, R6, and R8, fold enrichment was further compared between NIGT1.1 WT -MYC and NIGT1.1 L25A/ L39A -MYC by two-tailed Student's t-test, and P values are indicated.
https://doi.org/10.1371/journal.pgen.1009197.g004 monomer (Fig 4). To clarify the reason for this unexpected observation, we performed competition assays using non-labeled competitor DNAs harboring mutations at different NIGT1-binding sites within the P1 sequence that contained both non-palindromic and palindromic motifs ( Fig 5A). The competitor DNA harboring mutations in the palindromic motif, designated as non-labeled P1(m2), did not inhibit the interaction between labeled P1 probe and NIGT1.1 WT , while competitor DNAs with mutations in the non-palindromic motif, designated as non-labeled P1(m1), (m3), and (m1m3), competed potently with the labeled P1 probe ( Fig 5A). Hence, NIGT1.1 WT recognized only the palindromic motif in the P1 probe. Similar results were obtained using non-labeled mutant variants of the P2 probe ( Fig 5B).
The NIGT1.1 WT protein effectively bound to non-palindromic motifs in the P3 probe but not to non-palindromic motifs in the P1 probe. To resolve this conflicting result, binding of NIGT1.1 WT to a single non-palindromic motif was examined by EMSA using DNA sequences dissected from the P3 probe (P3-1 and P3-2), which contained only a single copy of the non-palindromic motif. The results showed that NIGT1.1 WT was unable to bind to P3-1 and P3-2 probes containing only one non-palindromic NIGT1-binding motif, whereas NIGT1.1 L25A/L39A could bind to these probes ( Fig 5C). The P1 probe contained two tandem repeats of the non-palindromic GAATC motif, whereas the P3 probe contained two copies of the non-palindromic GAATC motif in the reverse orientation. Thus, it was likely that each NIGT1.1 WT subunit of the NIGT1.1 WT dimer recognized one GAATC motif in the reverse orientation. To corroborate this hypothesis, competition assays were performed using nonlabeled mutant P3 probe in which either one or both non-palindromic NIGT1-binding sequences were mutated ( Fig 5D). Non-labeled competitor DNAs harboring mutations at either of these sites [P3(m1) and (m2)] or both sites [P3(m1m2)] competed less effectively with the labeled P3 probe than the wild-type competitor DNA. Consistent results were obtained in EMSAs using labeled P3(m1), (m2), and (m1m2) probes ( Fig 5E). Hence, we conclude that a single copy of the GAATC motif is not sufficient for the NIGT1 WT -DNA interaction; however, the NIGT1.1 dimer can bind to two copies of the GAATC motif in the reverse orientation, even if they are separated by a spacer sequence (a 38 bp sequence in probe P3), and then form a stable protein-DNA complex through the interaction of each subunit of the dimer with the GAATC motif. Thus, dual mode DNA recognition by NIGT1 proteins is not conferred by the recognition of GAATC and GAATATTC sequences but by the recognition of GAATATTC and GATTC-N 38 -GAATC sequences, revealing the uniqueness of dual mode DNA recognition by NIGT1 family proteins (described in further detail in the Discussion). On the other hand, NIGT1.1 L25A/L39A bound to a single copy of the GAATC motif with low affinity as a monomer.

Different modulations of the expression of nutrient response-associated genes by NIGT1.1 WT and NIGT1.1 L25A/L39A in vivo
Amino acid substitutions within the CCD of NIGT1.1 did not affect its nuclear localization ( Fig 2B). Furthermore, intact and truncated NIGT1.1 proteins fused to the GAL4 DNA-binding domain similarly repressed the GAL4-binding site-containing synthetic promoter (S4 Fig), indicating that NIGT1.1 represses transcription independently of the N-terminal CCD as long as NIGT1.1 can bind to target DNA sequences. Thus, NIGT1.1 L25A/L39A was assumed to possess no or reduced activity for repression of target gene promoters, due to its reduced affinity to target DNA sequences. Co-transfection assays using Arabidopsis protoplasts revealed that the NIGT1.1 L25A/L39A protein still repressed both SPX1 and NRT2.1 promoters; however, as hypothesized, the repression by the mutant NIGT1.1 L25A/L39A was weaker than that by NIGT1.1 WT (Fig 6A), consistent with the lower ability of NIGT1.1 L25A/L39A to bind to SPX1 and NRT2.1 promoters (Fig 4B). We note that the weaker repressor activity of NIGT1.1 L25A/L39A was not likely caused by its lower stability, since its half-life was not shorter than that of NIGT1.1 WT

PLOS GENETICS
The NIGT1 dimer mediates nutrient response in Arabidopsis assay. An effector plasmid used to express NIGT1.1 WT or NIGT1.1 L25A/L39A was co-transfected into Arabidopsis protoplasts together with a reporter plasmid harboring the LUC gene under the control of the SPX1 or NRT2.1 promoter and an internal control plasmid containing the GUS gene. LUC activity was normalized relative to GUS activity. Data represent mean ± standard deviation (SD; n = 4-5) of relative LUC activity. The value obtained with an empty effector plasmid (none) was set to 1 in each panel. Significant differences among values obtained with different effector plasmids were determined using one-way ANOVA, followed by Tukey's HSD test, and are indicated by different lowercase letters.  [30]. NIGT1.1 L25A/L39A suppressed SPX1 expression but to a lesser extent than NIGT1.1 WT ; by contrast, NIGT1.1 L25A/L39A and NIGT1.1 WT repressed the expression of NRT2.1 to similar levels ( Fig 6B). This is consistent with the observation that L25A and L39A amino acid substitutions in the CCD of NIGT1.1 exerted negative effects on NIGT1.1-mediated repression of SPX1 and NRT2.1 promoters to different extents; NIGT1.1 L25A/L39A retained only 44% of the repression activity of NIGT1.1 WT on the SPX1 promoter but 87% of the repression activity on the NRT2.1 promoter ( Fig 6A). Furthermore, since NIGT1 binding regions in the promoters of three nitrate transporter genes including NRT2.4, NRT2.5, and NAR2.1 had been identified by ChIP analysis [24,26], expression levels of these genes in nigtQ/NIGT1.1 WT and nigtQ/NIGT1.1 L25A/L39A plants were also examined. The result further clarified that diminishing the ability for dimerization exerts different effects on the expression levels of distinct target genes in planta ( Fig  6B). Interestingly, the result suggested that disruption of the CCD had no substantial effect on the NIGT1.1-meditated repression of the NRT2.1 and NAR2.1 promoters that lacked the palindromic GAAGATTC motif, while NIGT1.1 WT and NIGT1.1 L25A/L39A differentially repressed transcription from the SPX1, NRT2.4, and NRT2.5 promoters that contained at least one functional GAAGATTC motif.

Abolishing the NIGT1 dimerization compromises nutrient responses in planta
We previously showed that NIGT1 family proteins modulate Pi uptake and PSR via the repression of SPX1 [30]. Therefore, physiological significance of dimerization of NIGT1 family proteins was first examined through the analysis of Pi uptake and PSR-related phenotype of nigtQ/NIGT1.1 L25A/L39A transgenic lines. Since genes encoding Pi transporters that play a major role in Pi uptake from roots, PHT1;1 and PHT1;4 [42], are negatively regulated by SPX proteins during PSR [17], we analyzed the expression levels of these genes. The results showed that PHT1;1 and PHT1;4 transcript levels in nigtQ plants were lower than in nigtQ/NIGT1.1 WT lines but comparable to those in nigtQ/NIGT1.1 L25A/L39A lines (Fig 7A). Furthermore, similar to PHT1;1 and PHT1;4 transcript levels, expression levels of IPS1, a marker gene for PSR [5], were modified in nigtQ plants, nigtQ/NIGT1.1 WT , and nigtQ/NIGT1.1 L25A/L39A plants ( Fig  7A). These modifications were consistent with the reduced suppression of SPX1 expression in nigtQ/NIGT1.1 L25A/L39A lines (Fig 6B). Consistently, Pi uptake activity ( Fig 7B) and shoot Pi concentration (Fig 7C) in nigtQ/NIGT1.1 L25A/L39A lines were intermediate between nigtQ and nigtQ/NIGT1.1 WT lines. Next, the PSR of nigtQ/NIGT1.1 L25A/L39A lines was evaluated by Pi starvation-induced accumulation of anthocyanins (Fig 7D and 7E). Under Pi-deficient conditions, L25A and L39A amino acid mutations within the CCD of NIGT1.1 clearly attenuated transcription start sites, respectively. NIGT1 binding motifs are indicated. Significant differences between WT and nigtQ plants under each condition were determined using two-tailed Student's t-test. P values are indicated. Significant differences among the nigtQ, nigtQ/NIGT1.1 WT , and nigtQ/NIGT1.1 L25A/L39A lines were determined using one-way ANOVA, followed by Tukey's HSD test, and are indicated by different lowercase letters.
We also found that when grown on the soil, retardation of shoot growth and accumulation of Pi were evident in nigtQ/NIGT1.1 WT

Discussion
In the current study, we revealed that the domain containing a coiled-coil motif in the N-terminal region of NIGT1 family proteins mediates homologous and heterologous NIGT1-NIGT1 interactions, and plays a key role in dual mode recognition of cis-elements in target gene promoters by NIGT1 family proteins. Since NIGT1 family proteins are key regulators of nitrate response and PSR [24,[26][27][28][29][30], and the physiological significance of dimerization of NIGT1 family proteins was confirmed in the current study, the findings of this study are important for a deeper understanding of the regulation of nutrient responses in plants. Moreover, these findings also provide novel insights into how plant TFs evolutionarily gained the ability to recognize diverse DNA sequences.

Amino acid sequence of the CCD is highly conserved among NIGT1 family proteins
The CCD enables dual mode DNA recognition of cis-elements by NIGT1 family proteins and ensures proper regulation of the responses to nitrate and Pi. A search for the CCD sequence in the public database containing diverse genetic resources, namely, the Arabidopsis 1001 Genomes database (https://1001genomes.org/tools.html; last accessed in Jan 2020), which contains CCD sequences from 1,135 naturally occurring accessions [43], indicated that the core 22-aa sequence of the CCD is very highly conserved. Sequence conservation of the CCD is much higher than that of flanking regions and even higher than that of the GARP DNAbinding domain (S8 Fig). Furthermore, although amino acid substitutions were detected in these sequences at glutamate or lysine residues within the CCD, no substitutions of branched-chain amino acid residues were detected in the sequence that constitutes the hydrophobic surface of the coiled-coil motif. Similarly, analysis using the rice SNP-Seek database (https://snp-seek.irri.org/; last accessed in Jan 2020) to identify sequence variants in 3,024 rice accessions [44] revealed no amino acid substitutions within the CCD of OsNIGT1 (S8 Fig). These data suggest that the CCD of NIGT1 family proteins has been highly conserved during evolution, perhaps because CCD-mediated dimerization of NIGT1 family proteins is indispensable for some aspects of NIGT1-mediated transcriptional regulation and proper nutrient responses.

The N-terminal CCD confers a unique mode of DNA recognition to NIGT1 family proteins
Protein-protein interactions play important roles in the transduction of environmental signals and are important for proper function of enzymes and transporters. In some TF families, protein dimerization and/or oligomerization play crucial roles in the recognition of target DNA sequences and/or proper regulation of target genes [45][46][47][48]. For instance, in durum wheat (Triticum turgidum), a LATERAL ORGAN BOUNDARIES DOMAIN (LBD) family TF, TtRa2LD, forms a homodimer to increase the affinity for target cis-elements [47]. Similarly, dimerization of Arabidopsis AUXIN RESPONSE FACTORS (ARFs) enhances their affinity for target DNA sequences [46]. Furthermore, it was proposed that dimerization of eukaryotic TFs emerged during evolution to allow the recognition of a larger number of nucleotides [49]. In the NIGT1 family, dimerization is necessary for binding to two types of palindromic DNA sequences (i.e. GAATATTC and GATTC-N 38 -GAATC) with high affinity (Fig 8). In this study, we showed that NIGT1.1 WT could not bind to a single GAATC motif, whereas NIGT1.1 L25A/L39A could bind to this sequence (Fig 8) probably because the NIGT1.1 L25A/ L39A   Fig 8. Model displaying the possible modes of NIGT1-DNA interaction. NIGT1 binding to a single palindromic and two separate non-palindromic sites. Dimeric NIGT1 binds to the palindromic GAATATCC sequence with high affinity but not to a single copy of the GAATC sequences, while monomeric mutant NIGT1 binds to a single copy of the GAATC sequences with low affinity. Dimeric NIGT1 binds to the GATTC-N x -GAATC sequences with high affinity through recognition of a single GAATC motif by each subunit. Each subunit of the NIGT1 dimer is indicated in blue or brown. https://doi.org/10.1371/journal.pgen.1009197.g008

PLOS GENETICS
The NIGT1 dimer mediates nutrient response in Arabidopsis protein bares additional amino acid residues required for DNA recognition on the surface, which were buried within the interaction surface of the NIGT1.1 WT dimer, as hypothesized previously [32] (Fig 8); however, the affinity and specificity of NIGT1.1 L25A/L39A for a single GAATC motif did not appear to be high, because NIGT1.1 L25A/L39A did not bind to a single copy of GAATC sequence strongly and competitor DNAs did not compete with labeled DNA probes effectively in EMSA (Fig 5C). These data suggest that dimerization of NIGT1 family proteins is involved in both the modulation of specificity and increase in affinity for target DNA sequences through interaction with a larger number of nucleotides.
Consistent with the observation that the NIGT1.1 WT protein exclusively forms a dimer in solution, dual mode DNA recognition allowed NIGT1 family proteins to recognize two types of palindromic sequences, GAATATTC and GATTC-N 38 -GAATC in the NRT2.1 promoter, revealing the previously undetected recognition motifs. Some TFs, such as bacterial GabR and mammalian thyroid hormone receptors, bind as a dimer to cis-elements consisting of two sequences separated by a spacer [50][51][52]. Furthermore, binding of GabR TFs to the target DNA sequence is coupled with DNA bending because of the presence of a long (29 bp) spacer sequence. However, DNA binding of NIGT1 family proteins is distinguishable from that of GabR TFs. Binding of NIGT1 family proteins to the GATTC-N 38 -GAATC sequence needs to be coupled with a drastic conformational change in the NIGT1 dimer, because the orientation of the GAAT core sequence is opposite from those found in GAATATTC sequence. Repositioning of the DNA-binding domain, or existence of NIGT1 dimers in which the CCD of each monomer interacts in the opposite direction (i.e. parallel and antiparallel), as observed in CCDs of some proteins [53,54], may explain these phenomena. Furthermore, for binding to GATTC-N 38 -GAATC sequence, DNA bending might be necessary due to the presence of a large spacer sequence. Further studies are required to reveal precise mechanisms to provide new clues for resolving the confusing and/or seemingly conflicting results of DNA recognition by TFs.

Dimerization of NIGT1 is critical for the proper control of nutrient responses
The results of the current study indicated that the dimerization of NIGT1 family proteins is important for binding to palindromic NIGT1-binding sites in the SPX1 promoter and for modulating various aspects of PSR. The NIGT1 binding regions in the NRT2.4 and NRT2.5 gene promoters also contain the palindromic sequence for NIGT1 binding. These suggest that the dimerization ability of NIGT1 family proteins may have evolved to precisely bind to target gene promoters and to properly control nutrient responses.
The expression level of SPX1 was inversely proportional to the amount of NIGT1.1 binding to the NIGT1 binding sites in the SPX1 promoter. However, although the amount of NIG-T1.1 L25A/L39A binding to the authentic NIGT1 site in the NRT2.1 promoter was suggested to be smaller than that of NIGT1.1 WT binding to this site in vivo, NIGT1.1 WT and NIGT1.1 L25A/L39A suppressed NRT2.1 expression to a similar extent in planta (Figs 4C and 6B), causing a remaining question. A possible explanation for this phenomenon is that binding of a small amount of NIGT1 family proteins to this site is sufficient to adequately suppress transcription in the context of the NRT2.1 promoter in planta. Alternatively, NIGT1.1 L25A/L39A might artificially bind to GAATC motifs outside the region analyzed in ChIP assay in vivo and effectively repress NRT2.1 expression. This unexpected observation suggests that the loss of dimerization of NIGT1 family proteins may generate complex effects on the gene expression profile in planta, further emphasizing the importance of dimerization-mediated precise binding of NIGT1 family proteins on nutrient responses. Y2H assays revealed that NIGT1 family proteins are able to dimerize not only with other NIGT1 family proteins but also with some HHO family proteins. Nitrate-inducible genes encode NIGT1 family proteins, while non-nitrate-inducible genes encode HHO family proteins, suggesting that the physiological roles of HHO family proteins are different from those of NIGT family proteins. This is consistent with previous findings; HHO4 and HHO5 have been shown to modulate flowering time and meristem activity, respectively [55,56]. However, our recent results suggest that rice OsHHO3 and OsHHO4, close homologs of Arabidopsis HHO5 and HHO6, play central roles in N deficiency response in roots [10]. Considering that HHO5 and HHO6 interacted with only certain members of the NIGT1 family in this study, it is possible that homologous or heterologous dimer formation among NIGT1 and HHO family proteins is required for the regulation of some aspects of nutrient responses.
Although we focused on the role of the homologous NIGT1.1 dimer in the current study, formation of heterologous dimers comprising NIGT1 and HHO family proteins through interaction between the conserved CCDs may offer a more complex regulatory network for a variety of physiological processes. Untangling these complex networks will further our understanding of the mechanisms underlying nutrient signaling and will enable the design and implementation of desired nutrient responses.
To analyze gene expression and Pi concentration, seedlings were grown on vertically oriented agar plates containing modified half-strength Murashige and Skoog (1/2 MS) medium [58] supplemented with 0.5% sucrose, 3.5 mM MES (pH 5.7), and 0.8% agar. In the modified 1/2 MS medium, KH 2 PO 4 concentration was changed to 0 or 500 μM to produce Pi-replete and Pi-deficient conditions, respectively, and iron (Fe) concentration was also reduced to 2 μM to avoid excess Fe under Pi-deficient conditions [59][60][61]. Seedlings were grown at 23˚C under continuous illumination (60 μE m -2 s -1 ) for 9 d. For growth analysis using nitrate or ammonium as the sole N source, 5 mM KNO 3 or 5 mM NH 4 Cl was added to N-free 1/2 MS medium. When ammonium was used as the sole N source, 5 mM KCl was additionally added to provide potassium ion. To measure the anthocyanin content, five seeds were cultured in 1.5 mL of modified 1/10 MS solution containing 0.5% sucrose, 3.5 mM MES (pH 5.7), and 10 or 1,000 μM KH 2 PO 4 in a 24-well microtiter plate (#92424, TPP, Trasadingen, Switzerland). Seedlings were grown at 23˚C under continuous light (60 μE m -2 s -1 ) for 9 d, with occasional shaking. For analysis of growth on the soil, seeds were first germinated on 1/2 MS agar plates, and 1-w-old seedlings were transferred to a nutrient-containing peat (Jiffy 7, Sakata Seed Corp., Yokohama, Japan). The plants were grown under continuous illumination (100 μE m -2 s -1 ) at 23˚C for another 14 d before phenotypic analysis.

Y2H assay
Y2H assays were conducted using yeast (Saccharomyces cerevisiae) strain AH109, as described previously [64]. After transformation of AH109 cells with the plasmids of interest, transformants were selected on synthetic defined (SD) medium lacking Leu and tryptophan (Trp; SD/-Leu/-Trp) at 30˚C, and single colonies were streaked on plates containing SD/-Leu/-Trp or SD medium lacking Leu, Trp, His, and adenine (Ade; SD/-Leu/-Trp/-His/-Ade). Plates were photographed after 5 d of incubation at 30˚C.

Agroinfiltration and BiFC assay
Agrobacterium tumefaciens strain GV3101 (pMP90) was transformed with plasmids of interest and suspended in 0.2 mM acetosyringone. Leaves of 3-week-old N. benthamiana plants grown in soil were co-infiltrated, as described previously [65,66], with Agrobacterium carrying the plasmid of interest and Agrobacterium carrying the P19 silencing suppressor to suppress gene silencing [67,68]. Plasmids expressing the nGFP-or cGFP-GRC3 fusion [40] served as negative controls. GFP fluorescence derived from the introduced plasmids was observed using a fluorescence microscope (BX51; Olympus, Tokyo, Japan) equipped with a digital camera (DP80; Olympus) at 2 or 3 d post-agroinfiltration. Nuclei were stained with DAPI (Sigma, St. Louis, MO) infiltrated through the abaxial surface of the leaf prior to microscopy.

Recombinant protein purification, SEC-MALS, and EMSA
E. coli strain Origami B (DE3) (Merck Millipore) was transformed with pET32-NIGT1.1 and pET32-NIGT1.1 L25A/L39A plasmids. Transformants cultured overnight in LB medium were inoculated into fresh LB medium and further cultured until the OD 600 value reached 0.6. To induce recombinant protein production, the culture was treated with 1 mM isopropyl-β-Dthiogalactopyranoside (IPTG) at 15˚C for 18 h. Soluble protein was extracted from the transformants by sonication in extraction buffer (50 mM sodium phosphate buffer [pH 8.0], 0.1% Triton X-100, 500 mM NaCl, and 1× EDTA-free protease inhibitor [#11873580001, Merck Millipore]) supplemented with 1 mM imidazole. Recombinant His-tagged protein in the extract was bound to TALON Metal Affinity Resin (#8901-1; Takara Bio Inc.), followed by two washes with the extraction buffer supplemented with 45 mM imidazole. Finally, Trx-6×His-NIGT1.1-6×His protein was eluted using extraction buffer supplemented with 250 mM imidazole. The eluted protein was mixed with an equal volume of glycerol and stored at -20˚C until needed for further analysis.
EMSAs were performed using recombinant proteins as described previously [24]. DNAprotein binding was performed in a 15 μl reaction mixture containing 40 fmol of biotinlabeled probes in the presence or absence of molar excess of non-labeled competitor DNA, and electrophoresis was carried out using 4% polyacrylamide gels.

Co-transfection assays using Arabidopsis seedlings and protoplasts
To analyze the repression domain of NIGT1 family proteins (S4 Fig), co-transfection assays were performed by particle bombardment using 10-d-old Arabidopsis seedlings and a particle delivery system (PDU-1000/He; Bio-Rad, Hercules, CA), as described previously [26]. The reporter plasmid contained the gene for firefly luciferase (FLUC) under the control of the GAL4-binding site-containing synthetic promoter [70], whereas effector plasmids were designed to express the fusion of GFP or NIGT1.1 with the GAL4 DNA-binding domain under the control of the cauliflower mosaic virus (CaMV) 35S RNA promoter [26]. Additional effector plasmids expressing the N-terminal or C-terminal region of NIGT1.1 (1-48 and 49-344 aa, respectively) fused to the GAL4 DNA-binding domain were constructed using NIGT1.1 cDNA fragments. The plasmid containing the gene for Renilla luciferase (RLUC) under the control of the CaMV 35S RNA promoter served as an internal control [70]. Levels of FLUC and RLUC activity were measured using the Dual-Luciferase Reporter Assay System (Promega KK, Tokyo, Japan) using a microplate reader (Mithras LB940; Berthold, Bad Wildbad, Germany).
Co-transfection assays with Arabidopsis mesophyll protoplasts were performed as described previously [71]. Protoplasts were isolated from rosette leaves of 3-week-old wildtype Arabidopsis plants grown in 1/10 MS nutrient solution at 23˚C under continuous illumination (60 μE m -2 s -1 ). The reporter plasmids containing FLUC gene under the control of the SPX1 or NRT2.1 promoter were reported previously [24,30], and effector plasmids were designed to express NIGT1.1 WT -2×MYC or NIGT1.1 L25A/L39A -2×MYC under the control of the constitutive 35S-C4PPDK fusion promoter. A plasmid expressing 2×MYC under the control of the 35S-C4PPDK promoter served as a control. Another plasmid expressing GUS gene under the control of the UBIQUITIN10 (UBQ10) promoter was used as an internal control.

Gene expression analysis
Total RNA extraction, cDNA synthesis, and quantitative real-time PCR (RT-qPCR) using gene-specific primers (S1 Table) were carried out as described previously [30]. Quantification of gene expression was based on the standard curve method using the UBQ10 gene as an internal control.

Measurement of Pi and anthocyanin contents
Foliar Pi concentration in 9-d-old seedlings was determined using the molybdenum blue method, as described previously [30,73]. Briefly, Pi was extracted from frozen pulverized samples using 1% acetic acid. Ten microliters of the extract was mixed with 80 μL of 0.35% (NH 4 ) 2 MoO 4 solution containing 0.5 M H 2 SO 4 and 10 μL of 10% (w/v) ascorbic acid. After 1 h of incubation at 37˚C, absorbance at 820 nm was recorded using a microplate reader (Infinite M1000, TECAN Ltd., Zürich, Switzerland). Serial dilutions of NaH 2 PO 4 was prepared in 1% acetic acid and used to create a standard curve.
Anthocyanin content was determined using whole seedlings grown in liquid culture as described previously [30,40]. Briefly, five seedlings were immersed in 600 μL of 1% HCl prepared in methanol and shaken overnight at 4˚C. Subsequently, 300 μL of the extract was mixed with 200 μL each of water and chloroform. After a brief centrifugation, the absorbance of the upper layer was recorded at 530 and 657 nm using a spectrophotometer. Relative anthocyanin content was calculated using the following equation: where A 530 and A 657 represent the absorbance at 530 and 657 nm, respectively, and sample FW represents the fresh weight of whole seedlings (mg).

Measurement of nitrate uptake
Arabidopsis seedlings grown on agar plates containing 1/2 MS medium, 0.5% (w/v) sucrose, 0.8% agar, and 3 mM MES-KOH (pH 5.8) for 10 days were used for the experiment. Nitrate uptake was measured using 15 N-labeled nitrate, as described previously [24]. In brief, roots were submerged in 0.1 mM CaSO 4 for 1 min, and then incubated in 0.2 mM K 15 NO 3 for 5 min. After 1 min wash with 0.1 mM CaSO 4 , roots were completely dried. The N content and 15 N/ 14 N isotopic composition of roots were analyzed by SI Science Co., Ltd. (Yokohama, Japan).

Statistical analysis
Two-tailed Student's t-test was used to determine significant differences between wild type and nigtQ plants. To compare multiple genotypes including nigtQ/NIGT1.1 WT and nigtQ/ NIGT1.1 L25A/L39A lines, one-way analysis of variance (ANOVA) was conducted, followed by Tukey's Honest Significant Difference (HSD) post hoc test. In all comparisons, α = 0.05 was used as the significance threshold. . The reporter and effector plasmids were co-transfected into Arabidopsis leaves by particle bombardment. GFP fused to GAL4 BD was used as a control. FLUC activity was normalized relative to RLUC activity derived from an internal control plasmid [70], and the relative FLUC activity obtained using GAL4-GFP fusion was set to 1. Data represent mean ± standard deviation (SD; n = 4).  Fig 1A) and the GARP domain in each protein are indicated with blue and red squares, respectively. (DOCX) S1 Table. List of primers used in the study. (XLSX)