Ca2+-dependent GTPase, Extra-large G Protein 2 (XLG2), Promotes Activation of DNA-binding Protein Related to Vernalization 1 (RTV1), Leading to Activation of Floral Integrator Genes and Early Flowering in Arabidopsis*

Background: Extra-large G proteins, XLGs, are nuclear GTP-binding proteins with both Gα-like and novel domains. Results: Unlike general G proteins, XLG proteins are Ca2+-dependent GTPases, and XLG2 interacts with the nuclear protein RTV1 (related to vernalization 1). Conclusion: GTP-bound XLG2 promotes RTV1-chromatin interaction, leading to activation of floral integrator genes. Significance: XLGs unusually integrate Ca2+-based and G protein-based cellular pathways and control flowering. Heterotrimeric G proteins, consisting of Gα, Gβ, and Gγ subunits, play important roles in plant development and cell signaling. In Arabidopsis, in addition to one prototypical G protein α subunit, GPA1, there are three extra-large G proteins, XLG1, XLG2, and XLG3, of largely unknown function. Each extra-large G (XLG) protein has a C-terminal Gα-like region and a ∼400 amino acid N-terminal extension. Here we show that the three XLG proteins specifically bind and hydrolyze GTP, despite the fact that these plant-specific proteins lack key conserved amino acid residues important for GTP binding and hydrolysis of GTP in mammalian Gα proteins. Moreover, unlike other known Gα proteins, these activities require Ca2+ instead of Mg2+ as a cofactor. Yeast two-hybrid library screening and in vitro protein pull-down assays revealed that XLG2 interacts with the nuclear protein RTV1 (related to vernalization 1). Electrophoretic mobility shift assays show that RTV1 binds to DNA in vitro in a non-sequence-specific manner and that GTP-bound XLG2 promotes the DNA binding activity of RTV1. Overexpression of RTV1 results in early flowering. Combined overexpression of XLG2 and RTV1 enhances this early flowering phenotype and elevates expression of the floral pathway integrator genes, FT and SOC1, but does not repress expression of the floral repressor, FLC. Chromatin immunoprecipitation assays show that XLG2 increases RTV1 binding to FT and SOC1 promoters. Thus, a Ca2+-dependent G protein, XLG2, promotes RTV1 DNA binding activity for a subset of floral integrator genes and contributes to floral transition.

and ethylene. In addition, XLG3 is specifically involved in the regulation of root waving and skewing (21,22), and XLG2 interacts with the sole Arabidopsis canonical G␤ subunit, AGB1, and is involved in pathogen resistance (23). Expression analyses of XLG genes using RT-PCR and a reporter gene assay show that all three XLGs are widely expressed throughout the plant, suggesting that XLGs may have additional functions in plant development and signaling beyond those already reported (21).
In Arabidopsis, flowering time is controlled by multiple endogenous and environmental cues, including photoperiod, temperature, and gibberellins. Floral regulatory cues are integrated by a group of genes, including FT (flowering locus T), SOC1/AGL20 (suppressor of overexpression of constans 1), AGL24 (agamous-like 24), and LFY (leafy). These genes are often referred to as "floral integrators" because most known flowering pathways eventually regulate expression of these flowering genes (24 -27). Studies in Arabidopsis and rice have shown that the floral integrator FT is a major component of florigen (28,29). FT together with FD (flowering locus D) activates SOC1 in the meristem (30). Floral integrators, in turn, activate floral meristem identity genes, including AP1 (apetala 1) and FUL (fruitful) (30,31). Thus, there is an intricate regulatory network among floral integrators and floral meristem identity genes to initiate and maintain floral transition in response to endogenous and environmental cues (24,25,27,30,31).
Winter annual strains of Arabidopsis contain functional FRI (frigida) and FLC (flowering locus C) genes and flower extremely late unless plants are exposed to a prolonged period of cold (i.e. vernalization). FLC acts as a potent floral repressor by the direct suppression of floral regulators FT and SOC1 (32,33). Vernalization results in stable repression of FLC to allow rapid flowering (34). Similar to winter annual strains of Arabidopsis, a group of late flowering mutants flower late in an FLCdependent manner (35). These mutants are collectively referred as the autonomous pathway mutants, and vernalization results in rapid flowering in the autonomous mutants as well (35). Vernalization results in an increase in two repressive histone modifications of FLC chromatin: methylation of histone H3 Lys-9 and histone H3 Lys-27 (34,36). These vernalization-induced histone modifications are mediated by a number of proteins, including VRN2 (vernalization 2), VRN1 (vernalization 1), and VIN3 (vernalization-insensitive 3) (34,36). In this study, we describe the biochemical characterization of the XLG2 protein and address its function in regulation of RTV1 (related to vernalization 1) and in floral transition, thereby elucidating a role of this unusual G protein in the control of flowering time.

EXPERIMENTAL PROCEDURES
Plant Materials and Vernalization Treatment-Seeds were germinated on 0.8% agar plates with half-strength Murashige and Skoog salt, 0.8% sucrose, stratified at 4°C for 2 days, and grown for 10 days in long day conditions (16 h of light at 22°C, and 8 h of dark at 22°C) or short day conditions (8 h of light at 22°C, 16 h of dark at 22°C), and then plants were transplanted to soil or harvested for RNA isolation. Light intensity for LD 2 and SD conditions was ϳ70 mol m Ϫ2 s Ϫ1 . For the vernalization treatment, seeds were germinated on agar plates for 5 days in SD conditions, and vernalized for 5 weeks at 4°C in SD (37), and then plants were harvested for RNA isolation, or plants were transplanted to soil and transferred to growth chambers (21 Ϯ 2°C) under LD or SD conditions for flowering time tests.
Protein Expression in Escherichia coli and Generation of XLG2(T475N) Mutant Construct-For expression in E. coli, the coding regions of the XLG genes and GPA1 cDNA were subcloned into the pET28a-modified expression vector (Novagen) which results in production of fusion proteins with N-terminal His 6 and thioredoxin (TRX) tags. We fused XLG proteins with TRX to increase solubility of the recombinant proteins. The coding region of RTV1 and each XLG was also subcloned into the pGEX-2T vector, which produces GST fusion proteins. XLG C-terminal regions (amino acids 486 -888 for XLG1, 461-862 for XLG2, and 433-849 for XLG3) were also subcloned into the same vector. All constructs were transformed into the E. coli BL21(DE3)pLysS strain. Transformed cells were inoculated into 10 ml of LB medium containing 25 g/ml kanamycin for pET28a and 50 g/ml ampicilin for pGEX-2T. The cultures were grown at 30°C overnight and then poured into 500 ml of fresh LB medium. When an A 600 of 0.8 was reached, isopropyl ␤-D-thiogalactopyranoside was added to the cultures to a final concentration of 0.9 mM, and the incubation was continued for another 4 h at 30°C for induction. His-tagged fusion proteins were purified using Ni 2ϩ -nitrilotriacetic acid-agarose column chromatography (Qiagen) (supplemental Fig. 1, a and  b), and GST fusion proteins were purified using glutathioneagarose (Sigma) according to each manufacturer's instructions. Empty vectors were transformed into the E. coli BL21 (DE3) pLysS strain and expressed for purification of His 6 TRX or GST. To purify XLG proteins lacking any tag, GST was cleaved by thrombin (Sigma) from GST fusion XLG proteins. Mutant (T475N) forms of full-length XLG2 and XLG2C were constructed by PCR using the following primers: 5Ј-GGTGGCGC-CAATACAATTTACAAA-3Ј and 5Ј-TTTGTAAATTGTA-TTGGCGCCACC-3Ј.
GTP Binding Assay-The [ 35 S]GTP␥S binding activity of full length XLGs, C-terminal regions of the XLGs, GPA1, and TRX protein was measured by the rapid filtration method (38,39). A 100 nM concentration of each purified recombinant proteins was incubated at 30°C in 200 l of 50 mM Tris-Cl buffer, pH 8.0, containing 1 mM EDTA, 1 mM dithiothreitol (DTT), 1 mM NaN 3 , 0.2 M [ 35 S]GTP␥S, and 10 mM MgCl 2 or CaCl 2 . The reaction was stopped by the addition of 2 ml of ice-cold buffer (20 mM Tris-Cl, pH 8.0, 100 mM NaCl, and 25 mM MgCl 2 ). Aliquots (20 l) were taken at the indicated times and vacuumfiltered onto nitrocellulose membranes (Millipore, Billerica, MA), which had been rinsed with 3 ml of ice-cold buffer, and membrane-bound material was washed rapidly three times with 3 ml of ice-cold buffer. The washed membranes were dried and quantified for radioactivity by liquid scintillation counting.
GTPase Assay-GTP hydrolysis assays were performed as previously described (40) using PEI-cellulose TLC plates. Briefly, the reaction was performed at 30°C in 200 l of buffer containing 50 mM Tris-Cl, pH 8.0, 1 mM EDTA, 1 mM dithiothreitol, 1 mM NaN 3 , 20 pmol of [␣-32 P]GTP, 100 nM full-length or C-terminal XLG protein or 100 nM of GPA1 protein, and 10 mM MgCl 2 , MnCl 2 or CaCl 2 . Aliquots of 10 l were sampled at appropriate time intervals, and 10 l of 0.5 M EDTA (pH 8.0) was added to the aliquots to stop the reaction. Samples (2 l) were spotted onto PEI-cellulose TLC plates, which were then developed in 0.5 M KH 2 PO 4 , pH 3.4, solution. After drying, the plates were exposed to a storage phosphor screen (Amersham Biosciences) and then scanned by a PhosphorImager (GE Healthcare).
Yeast Two-hybrid Screening-Yeast two-hybrid screening was carried out according to the manufacturer's protocol (Clontech). Briefly, the bait plasmid, pBUTE-XLG2, and total cDNAs of Arabidopsis obtained from ABRC (catalogue number CD4-10) in which the cDNAs were fused to the pACT-AD were used sequentially to transform Saccharomyces cerevisiae strain pJG69-4A. The resulting transformants were then plated on synthetic complete (SC) medium lacking uracil (Ura), leucine (Leu), and histidine (His) (SCϪ, Ura Ϫ , Leu Ϫ , and His Ϫ ) and/or SCϪ Ura Ϫ and Leu Ϫ , including 3 mM 3-aminotriazole. Histidine-positive colonies were further tested for ␤-galactosidase (LacZ) activity according to the manufacturer's protocol (Clontech). The clones of cDNAs encoding candidate XLG2 interacting proteins were selected and isolated based on their growth as blue colonies in histidine-deficient medium. The cDNAs were isolated and sequenced. The DNA sequences were then subjected to BLAST analysis using the GenBank TM database.
In Vitro Protein Pull-down Assay-Purified GST or GST-RTV1 was incubated for 2 h at 4°C with 40 l of glutathione-Sepharose in PBST (PBS containing 1% Triton X-100). After three washes, the beads were incubated with His-XLG2C in 0.5 ml of PBST for 3 h on a rotary shaker at 4°C. The beads were then washed five times with PBST, boiled with SDS sample buffer, and subjected to 12% SDS-PAGE. Western blot of the resolved proteins was carried out using anti-His (Qiagen catalog no. 34660) and anti-GST (Novagen catalog no. 71097) monoclonal antibodies.
Transient Expression Proteins Fused with Green Fluorescent Protein-RTV1 cDNA was fused upstream of the GFP cDNA under the control of the CaMV 35S promoter (pUC::GFP) (41). The construct (RTV1-GFP) was co-transformed into Arabidopsis protoplasts along with the red fluorescent protein cDNA fused to a nuclear localization signal peptide (NLS-RFP (42)). Transient expression of GFP-and RFP-fused constructs in Arabidopsis protoplasts was performed according to the method described by Heo et al. (40). Briefly, recombinant plasmids were introduced by polyethylene glycol (PEG)-mediated transformation into Arabidopsis protoplasts that had been prepared from leaf tissues. Expression of the fusion constructs was monitored after transformation and observed by fluorescence microscopy (Olympus AX70 TR, Olympus).
Production of Transgenic Arabidopsis Lines-For generation of P RTV1 ::GUS lines (Col background), the RTV1 promoter (0.91 kb, which is the complete intergenic region) was amplified and subcloned into the pENTR directional TOPO vector (Invitrogen) followed by LR reaction (Invitrogen) into the destination vector pHGWFS7 (43) using LR enzyme (Invitrogen). GUS activity was revealed by incubation in 100 mM NaPO 4 (pH 7.2), 2.5 mM 5-bromo-4-chloro-3-indoyl-␤-D-glucuronide. Plant tissue was incubated at 37°C for 24 h. After staining, chlorophyll was cleared from the samples by dehydration using 70% ethanol.
To generate 35S-RTV1 lines, RTV1 cDNA was ligated to the pH7GWIWG2 vector, and the construct was transformed into wild-type plants (Col-0) via Agrobacterium-mediated transformation (44). For 35S-XLG2 RTV1, XLG2 cDNA was ligated to pGWB11, and the resulting 35S-XLG2 construct was transformed into the T1 generation of 35S-RTV1 (13) plants. Transgenic lines were selected on selective antibiotic marker-containing medium and were confirmed by genotyping PCR. To generate 35S-RTV1-GFP ft-1 lines, the 35S-RTV1-GFP construct was transformed into ft mutant plants by floral dipping, and then T1 generation plants were selected by selective antibiotic-containing medium, followed by flowering time determination.
Electrophoretic Mobility Shift Assay (EMSA)-EMSAs were performed as described by Despres et al. (45). DNA fragments of FLC, SOC1, FT, VRN2, and DFR promoters 400 bp upstream of their start codons were generated by PCR amplification and confirmed by sequencing. 100 ng of double-stranded synthetic nucleotides were radiolabeled using T4 polynucleotide kinase. EMSA reactions were performed at room temperature for 30 min in 15 l of EMSA buffer (20 mM Hepes-KOH, pH 7.9, 100 mM KCl, 2 mM DTT, and 20% glycerol) containing 1 l (20,000 cpm) of the labeled nucleotides and 0 -2 g of RTV1. In addition, 100 nM XLG2C, 10 mM Ca 2ϩ , and 1 mM GTP␥S were added with the buffer in some reactions. EMSA reactions were loaded onto a 5% Tris-borate EDTA-polyacrylamide gel and run at 8 V/cm for 90 min. After electrophoresis, the gel was dried and subjected to autoradiography at Ϫ80°C using Eastman Kodak Co. x-ray film.
Chromatin Immunoprecipitation (ChIP) Assays and Immunoprecipitation-ChIP assays were carried out (Johnson and Bresnick (46)) as described by Heo et al. (47), following the protocol of Gendrel et al. (48). Chromatin proteins and DNA were cross-linked in 10-day-old Col, 35S-RTV1 (13) and 35S-XLG2(1) RTV1 seedlings by formaldehyde fixation. Chromatin was isolated using the same extraction buffers as in Heo and Sung (47) and Gendrel et al. (48). Chromatin size postsonication was 500 -800 bp, as confirmed by agarose gel electrophoresis (not shown). Chromatin immunoprecipitation was performed using anti-FLAG (Sigma; catalog no. F1804) or anti-GFP (Molecular Probes; catalog no. A11122) antibodies. Crosslinks were reversed by incubation at 65°C for 6 h, and DNA was purified with QIAquick spin columns (Qiagen) and eluted in 50 l of Tris-HCl EDTA (pH 8.0). PCR was used to amplify fragments of the FT, SOC1, and FLC promoters as used in EMSAs. The relative -fold increase of retrieved DNAs was calculated as follows: first, relative enrichment of each fragment was calcu-lated by comparison with the input (chromatin sample before immunoprecipitation). Second, the background levels of retrieved DNAs were calculated by measuring ACTIN2 enrichment compared with input. Third, relative levels of each retrieved fragment were calculated using the formula, C T ϭ C T (each gene) Ϫ C T (ACTIN2). Last, relative levels in Col were set equal to 1 in all cases, and -fold change relative to Col was calculated. Primers are given in supplemental Table 1.

GTP Binding and Hydrolytic Activities of XLG2 Protein-The
C-terminal regions of XLGs contain G␣ domains (G-1 to G-5) (20,21), which, in heterotrimeric G proteins, mediate GTP binding and hydrolysis (49). However, some amino acids conserved in the G-1 to G-5 domains of G␣ proteins are not conserved within the corresponding domains of XLG proteins. Particularly, amino acids in the regions of the G-3 and G-5 domains, which are involved in Mg 2ϩ coordination, GTP binding, and subsequent GTP hydrolysis, are not conserved (50) ( Table 1). This raised the question of whether or not XLG proteins are able to bind and hydrolyze GTP. We addressed the biochemical properties of the three XLG proteins using GTP binding and GTPase assays. Results were similar for all three XLG proteins; results for XLG2 are reported here, and those for XLG1 and XLG3 are described in supplemental Figs. 1-4. To examine specific GTP␥S binding and GTPase activities, we purified and assayed recombinant GPA1 and XLG proteins fused with His 6 and TRX.
As expected, GPA1 bound GTP␥S rapidly in the presence of Mg 2ϩ (Fig. 1A), illustrating the known dependence of G␣ proteins on this divalent cation (49). However, XLG2C did not bind GTP␥S under these conditions (Fig. 1A). To address whether XLG2C has a preference for other divalent cations over Mg 2ϩ for GTP binding activity, GTP binding filter assays were performed in the presence of Ca 2ϩ instead of Mg 2ϩ . Interestingly, the presence of Ca 2ϩ promoted XLG2C binding of GTP␥S, whereas in the absence of any divalent ion, there was no GTP␥S binding to XLG2C (Fig. 1B). By contrast, GPA1 showed weaker GTP␥S binding in the presence of Ca 2ϩ as compared with Mg 2ϩ (Fig. 1C). Similar to XLG2C, XLG1 and XLG3 also bound GTP␥S in the presence of Ca 2ϩ (supplemental Figs. 3 and 4). These results indicate that XLG proteins prefer Ca 2ϩ over Mg 2ϩ as a cofactor for their GTP binding activities. To confirm the specificity of GTP binding, PCR-based mutagenesis was used to generate a mutation, T475N, in the G-1 domain of XLG2, a residue which is essential for binding to the phosphate moieties of guanine nucleotides (51). Similar mutations have been reported to abolish the GTP binding activity of other G proteins. For example, the T22N mutation of rice OsRab7 protein eliminates GTP binding (52,53). Mutations S47C of mammalian G␣ o and S48C of G␣ i2 also impair GTP␥S binding and create a dominant negative phenotype (54,55). We designated this mutant protein XLG2C(T475N); this protein was also tagged with N-terminal His 6 and TRX. GTP binding by XLG2C(T475N) protein was greatly reduced (Fig. 1C). As expected, purified His 6 TRX control protein did not show any GTP binding activity (Fig. 1, A and C). Analogous mutations reduced the GTP binding of XLG1C and XLG3C (supplemental Figs. 3A and 4A).
Because heterotrimeric G protein ␣ subunits, including Arabidopsis GPA1, have intrinsic GTPase activity, we examined the GTPase activity of XLG2C using thin layer chromatography (TLC) in the presence of various divalent cations. Although GPA1, like mammalian G␣ subunits, exhibited the greatest hydrolytic activity in the presence of Mg 2ϩ , XLG2C showed little or no GTPase activity under this condition ( Fig. 2A). In contrast, XLG2C GTPase activity was observed in the presence of Ca 2ϩ (Fig. 2, A and B), although this activity was less than that of GPA1 in the presence of Mg 2ϩ ( Fig. 2A and supplemental Fig.  2C). As expected, no GTPase activity was observed for the mutant form of XLG2C, XLG2C(T475N) (Fig. 2B). The GTPase activity of XLG2C protein, which has no tag, was similar to that of His 6 TRX-tagged XLG2C (supplemental Fig. 2B), indicating that His 6 TRX tag does not affect the GTP binding and GTPase activities of XLG C terminus proteins. The specificity of the GTPase activity of XLG2C was also addressed (Fig. 2C). When a ϳ100-fold excess of cold GTP␥S, GDP, or ATP was added to the reaction, GTP␥S had the greatest competitive effect, indicating that XLG2C specifically binds GTP (Fig. 2C). The full-length XLG2 protein exhibited only slightly increased GTPase activity compared with XLG2 C terminus protein ( RTV1 Is Identified as an XLG2-interacting Protein by Yeast Two-hybrid Screen-To address the functions of XLGs, we used a yeast two-hybrid screen to identify proteins that interact with XLG2. One XLG2-interacting protein recovered from the screen was RTV1 (related to vernalization 1) (At1g49480). RTV1 encodes a predicted protein of 226 amino acids with a plant-specific B3 DNA binding domain and a nuclear localization signal (NLS) (56). RTV1 shows an overall 81% identity with VRN1 in its amino acid sequence. VRN1 is a plant-specific B3-like DNA binding domain protein (57) that is required for maintenance of vernalization-mediated FLC repression and also promotes flowering in a vernalization-independent manner (58). To identify the domains of XLG2 that are responsible for RTV1 binding, we used partial cDNAs for additional yeast two-hybrid assays (Fig. 3A). Both full-length XLG2 and the C terminus of XLG2 interacted with RTV1, but the N terminus of XLG2 did not, indicating that the C terminus of XLG2 is responsible for the interaction in yeast.
Interestingly, the XLG2 mutant proteins, XLG2 full-length (T475N) and XLG2C(T475N), which do not bind GTP, did not show any interaction with RTV1 in yeast (Fig. 3A), suggesting that XLG2 interacts with RTV1 in a GTP-dependent manner. These results were further confirmed by in vitro pull-down assays using recombinant proteins. We found that His 6 TRX XLG2C but not His 6 TRX XLG2C(T475N) binds to GST-RTV1 in vitro (Fig. 3B). Thus, the GTP binding activity of XLG2 appears to be required for its interaction with RTV1. Given the sequence similarity between RTV1 and VRN1, we also tested whether VRN1 interacts with XLGs using the yeast two-hybrid assay. However, VRN1 did not interact with any of the three XLGs (data not shown). GTPase activity was measured using the PEI-cellulose TLC plate method. B, GTPase activity of XLG2 and lack of GTPase activity of XLG2C(T475N) in the presence of CaCl 2 . The reactions were performed as described in A. Aliquots (2 l) were withdrawn after incubation for 2 h and then separated by TLC. C, substrate competition test for XLG2 in the presence of CaCl 2 . Reactions were performed as described in A in the presence of competitors (3 M unlabeled GTP␥S, GDP, or ATP). The graph was drawn using quantitative TLC results (supplemental Fig. 2A) measured by ImageJ software. GTP hydrolysis was calculated as GDP/(GTP ϩ GDP). GTP hydrolysis in the absence of competitor was designated as 100%. Error bars, S.D.

TABLE 2 Comparison of GTPase activities of XLG proteins
GTPase activity calculated as the fraction of GTP remaining (GTP/(GTP ϩ GDP)) after a 2-h incubation, as quantified by TLC. Smaller values indicate greater activity. Data shown are means Ϯ S.D. of three independent replicates. ND, not determined. FL, full-length XLG proteins. C, C termini of XLG proteins, which include GTP binding domains.  (T475N) were fused to the bait (BD) vector, and a full-length cDNA of RTV1 was fused to the prey (AD) vector. These constructs (described at the right) were co-transformed into pJG69-4A yeast cells, and their interactions were evaluated. AGG1 and AGB1 were used as a positive control for interaction (14). B, in vitro protein pull-down assay. Approximately 1 g of GST or recombinant GST-tagged RTV1 was incubated with 1 g of recombinant His-tagged AtXLG2C or AtXLG2C(T475N) and pulled down with GST resin after preblocking with 1% BSA. After five washes, the bound protein was fractionated by 12% SDS-PAGE and subjected to Western blot analysis with anti-GST or anti-His monoclonal antibodies. The extra bands may represent nonspecific cross-reactivity to bacterial proteins. Detection was by alkaline phosphatase/NBT directly on the blot membrane (left) or by enhanced chemiluminescence on x-ray film (right). (Fig.  4). Total proteins were extracted from wild type (non-transgenic), 35S-FLAG-XLG2(1), and 35S-FLAG-XLG2(1) 35S-RTV1-GFP double-overexpressing lines. Protein extracts were immunoprecipitated with anti-FLAG antibody as probe, and then protein gel blotting was performed to detect RTV1-GFP with anti-GFP antibody (Fig. 4). RTV1-GFP was detected in the proteins from 35S::FLAG-XLG2(1) 35S-RTV1-GFP doubleoverexpressing plants immunoprecipitated with anti-FLAG antibody. In control experiments, RTV1-GFP was not detected in the proteins immunoprecipitated with anti-FLAG antibody from WT or from 35S-FLAG-XLG2(1) plants, confirming the physical interaction of XLG2 with RTV1 in plant cells.

XLG2 and RTV1 Interact in Vivo and RTV1 Protein Localizes to Nucleus-We used co-immunoprecipitation to test the protein-protein interaction between XLG2 and RTV1 in vivo
Given that XLG2 and RTV1 interact in planta, they would be expected to show identical or overlapping subcellular localization. All XLG proteins, including XLG2, have a nuclear localization sequence and localize to the nucleus, as observed in transient expression assays (21). We carried out transient expression assays to ascertain the localization of RTV1, which also has a predicted nuclear localization sequence (RKKK; amino acids 69 -72). The green fluorescence signal from RTV1-GFP overlaps with the red fluorescence from the RFP-tagged NLS peptide, indicating that RTV1 localizes to the nucleus (Fig. 5).
RTV1 Is Widely Expressed-We applied several complementary approaches to assess RTV1 expression patterns. RT-PCR analysis showed that, similarly to XLG2 (21), RTV1 is expressed throughout the plant, including in roots, stems, leaves, flowers, and cauline leaves (Fig. 6A). We then used reporter gene analysis to examine expression patterns in more detail. Histochemical GUS staining of P RTV1 ::GUS lines showed that RTV1 is strongly expressed in cotyledon (a), shoot apical meristem (b), radicle (c), root tip (d), and staminar filaments (e) (Fig. 6B).
Given the sequence similarity between VRN1 and RTV1 and thus a possible role of RTV1 in vernalization response, we also examined mRNA expression of RTV1 during the course of ver-nalization treatment (Fig. 6, C and D, and supplemental Fig. 5). RNA blot analysis showed that mRNA expression of RTV1, but not of XLG2, is increased after a 5-week cold treatment (Fig.  6C). Up-regulation of RTV1 by a 5-week cold treatment was also observed using P RTV1 -GUS (Fig. 6D). Stronger expression of P RTV1 -GUS was observed in plants that experienced a long term cold treatment (Fig. 6D, ϩV) compared with a control without cold treatment (Fig. 6D, ϪV).
To further dissect the increased level of RTV1 mRNA, we performed quantitative real-time RT-PCR using samples subjected to cold treatments ranging from 1 h to 5 weeks    MARCH 9, 2012 • VOLUME 287 • NUMBER 11 (supplemental Fig. 5). The mRNA level of RTV1 was rapidly increased almost 10-fold within 1 h of cold treatment and then decreased below the level of no cold treatment after 3 h of cold treatment. RTV1 mRNA expression eventually increased again ϳ2-fold after a 5-week cold treatment (supplemental Fig. 5). XLG2 mRNA also exhibited the rapid increase within 1 h of cold treatment and the decrease after 3 h of cold treatment. Unlike RTV1, however, the level of XLG2 mRNA remained unchanged after a 5-week cold treatment ( Fig. 6C and supplemental Fig. 5).

XLG2 and RTV1 Overexpression Promote Flowering
RTV1 Overexpression Promotes Early Flowering, Which Is Enhanced by Co-overexpression of XLG2-The ectopic expression of VRN1 causes early flowering due to the activation of FT and SOC1 (58). We accordingly examined flowering times of transgenic plants overexpressing RTV1. The 35S-FLAG-RTV1 plants exhibited rapid flowering under SD and a slight acceleration of flowering under LD (Fig. 7, A and B). In addition, when 35S-FLAG-RTV1 plants were vernalized, flowering was further accelerated (Fig. 7, A and B). Because XLG2 and RTV1 interact, we also evaluated the flowering time of 35S-FLAG-XLG2 35S-RTV1-GFP double overexpression lines. Interestingly, double overexpressor (Fig. 7, A and B, 35S-FLAG-XLG2 35S-RTV1-GFP) plants showed earlier flowering under LD and SD compared with both WT and 35S-FLAG-RTV1 plants, with or without vernalization treatment (Fig. 7B). Thus, the interaction between XLG2 and RTV1 appears to enhance RTV1 activity in vivo. To investigate the molecular basis of this early flowering phenotype, we examined the expression of floral pathway genes in 35S-FLAG-XLG2 and 35S-FLAG-RTV1 overexpression lines and in 35S-FLAG-XLG2 35S-RTV1-GFP double overexpres-sion lines under SD using quantitative RT-PCR. Introduction of 35S-FLAG-XLG2 alone did not promote transcript levels of the floral pathway genes SOC1, AGL24, FT, and LFY (Fig. 7C). However, the abundance of these transcripts was increased in 35S-FLAG-RTV1 plants and even more so in 35S-FLAG-XLG2 35S-RTV1-GFP plants (Fig. 7C).
Vernalization triggers the repression of FLC and other FLC clade genes (37,59). mRNA levels of FLC and its clade genes MAF1, MAF2, and MAF5 in 35S-FLAG-RTV1 and 35S-FLAG-XLG2 35S-RTV1-GFP plants immediately after vernalization showed no significant difference compared with vernalized wild type in the extent of this repression (i.e. -fold change compared with Col was Ͻ2). After vernalization, however, the floral pathway integrator genes SOC1 and AGL24 and particularly the levels of FT mRNA were higher in these lines than in Col (i.e. -fold change compared with Col was Ͼ2) (Fig. 7D). To analyze the effect of ft mutation (60) on the flowering time of 35S-RTV1-GFP plants, 35S-RTV1-GFP construct was transformed into the ft mutant, and flowering times were determined (Fig.  8). 35S-RTV1-GFP in the ft mutant background had the same late flowering phenotype as ft mutants (Fig. 8), indicating that early flowering of 35S::RTV1 requires the elevated expression of FT. Therefore, accelerated flowering of overexpressing XLG2 and RTV1 plants (with or without vernalization) apparently occurs directly through regulation of the floral pathway integrator gene, FT, through a mechanism independent of repression of FLC and its clade genes.
RTV1 Binds DNA and XLG2 Enhances This DNA Binding Activity-VRN1 shows a non-sequence-specific DNA binding activity in vitro (58). Given the sequence similarity between VRN1 and RTV1 and the nuclear localization of RTV1, we examined the DNA binding activity of recombinant RTV1 protein. EMSAs were performed using DNA fragments derived from promoters of FLC, SOC1, FT, VRN2, or DFR (dihydroflavonol 4-reductase). Recombinant RTV1 protein bound to all DNA fragments tested ( Fig. 9A and supplemental Fig. 6). These DNA fragments do not share any common sequence motif, illustrating that, similar to VRN1, RTV1 has non-sequencespecific DNA binding activity, at least in these in vitro assays. The interaction between XLG2 and RTV1 prompted us to hypothesize that this interaction may modulate the DNA binding activity of RTV1. Accordingly, recombinant XLG2 C-terminal domain protein (XLG2C), which contains the RTV1 interaction domains (Fig. 3), was added in the reaction with RTV1, and the binding strength was evaluated by the amount of probe retardation (Fig. 9B). The addition of XLG2C increased the DNA binding activity of RTV1 (Fig. 9B), whereas the mutant protein XLG2C(T475N), which fails to interact with RTV1, did not (Fig. 8B). Because XLG2C(T475N) mutant protein also fails to bind GTP in vitro (Fig. 1C), we examined the effect of GTP binding of XLG2C on the DNA binding activity of RTV1. XLG2C, GTP␥S, and Ca 2ϩ were added in the reaction together with RTV1. Under these conditions, XLG2C formed a supercomplex with RTV1 and probes, suggesting that activation of XLG2C by GTP binding contributes to the enhanced DNA binding of RTV1 in vitro (Fig. 9C).

RTV1 Is Associated with FT and SOC1 Promoters in Vivo-
We showed that RTV1 overexpression lines show early flowering due to increased expression of floral pathway integrator genes independent of regulation of FLC expression (Figs. 7 and 8) and that RTV1 binds to DNA in a non-sequence-specific manner in vitro (Fig. 9A and supplemental Fig. 9). To address whether RTV1 binds to promoters of floral pathway integrator genes, such as FT and SOC1, in vivo, ChIP assays were employed. For these assays, wild type, 35S-FLAG-RTV1 (13), and 35S-FLAG-XLG2(1) 35S-RTV1-GFP double-overexpresssing lines were used. After immunoprecipitation with anti-FLAG or anti-GFP antibodies, enrichment of FT and SOC1 pro-moter regions was assayed by quantitative PCR. As shown in Fig. 10, B and C, enrichment of the FT promoter region, 1-200 bp upstream from the start codon (P1) was observed in the 35S-FLAG-RTV1(13) line, with slightly enhanced enrichment in the 35S-FLAG-XLG2(1) 35S-RTV1-GFP overexpression lines. In addition, SOC1 promoter regions were not enriched in 35S-FLAG-RTV1 (13), whereas enrichment of the SOC1 promoter region, 1-200 bp upstream from the start codon (P1) was observed from the 35S-FLAG-XLG2(1) 35S-RTV1-GFP (Fig.  10C). No enrichment of FLC regions was observed in either the 35S-FLAG-RTV1 (13) line or the 35S-FLAG-XLG2(1) 35S-RTV1-GFP line (Fig. 10, B and C). Taken together, our results indicate that XLG2 promotes the DNA binding activity of RTV1 specifically to promoter regions of FT and SOC1 in vivo and thus leads to the activation of floral integrator genes (Fig.  11).

DISCUSSION
XLG Proteins Are Specific GTPases-Here we report new biochemical and functional properties of the XLG proteins. Conventional G␣ proteins have highly conserved domains involved in GTP binding and hydrolysis. Because of variations in amino acid sequences within these domains in the three XLG proteins (Table 1), it was not clear whether or not the XLG proteins function as GTPases. XLG1 had been shown previously to bind GTP (20), but GTP binding of XLG2 and XLG3, or GTPase activity of any of the XLG proteins had not been evaluated. In this study, we show that the three XLG proteins indeed specifically bind GTP and possess GTPase activity. Thus, we provide biochemical evidence that the XLG proteins, despite their unusual structure, are bona fide G proteins. Given the variations in amino acid sequences of the XLG GTP-binding domains, we initially hypothesized that the XLG-specific N termini might provide structural scaffolds that enable GTP binding. However, this appears not to be the case because the C termini of the XLGs alone are sufficient for GTP binding and GTPase activity ( Fig. 1 and supplemental Figs. 2-4). The specific functions of the unique N termini of the XLG proteins remain to be uncovered. XLG2, which is the most divergent of the three XLGs in the G-1, G-2, G-3, and G-4 domains, also has the slowest GTPase activity ( Table 2). Other types of G proteins, such as Ras and EF-Tu (61), associate with GTPase-activating proteins (GAPs) (62). It will be interesting in the future to determine whether GAPs exist for the XLGs.
The XLGs are distinct from canonical G␣ subunits in their dependence on Ca 2ϩ , rather than Mg 2ϩ , as a cofactor. The G-1 to G-3 domains of G proteins provide critical contacts for the ␣-, ␤-, and ␥-phosphates of the guanine nucleotide and also for the coordination of Mg 2ϩ (50). The G-4 and G-5 domains are important for the binding of the guanine ring (50). XLG2 has the conserved amino acid motifs in the G-1, G-2, and G-4 regions. XLG2 utilizes Ca 2ϩ but not Mg 2ϩ as a cofactor although it has the well conserved threonine residue in the G-2 region (Table 1), which plays a critical role in coordinating Mg 2ϩ (50). However, XLG proteins do not have the conserved motifs of the G-3 and G-5 regions (21) ( Table 1). The glutamine residue in the G-3 (DXXGQ) domain is involved in coordination of Mg 2ϩ mediated by water, and the main chain of the glycine residue forms hydrogen bonds with the ␥-phosphate of GTP (49,51). When introduced in Ras, virtually all mutations in the G-3 domain reduce its GTPase activity and disrupt Ras binding to Ras-GAP (63). Corresponding mutations in G␣ s (64,65) and in G␣ i1 (66) also reduce the GTPase activity. In addition, mutation at the G-5 (TCA) alanine in G␣ i promotes rapid GDP dissociation, although the intrinsic GTPase activity of G␣ i is not changed (67). Taken together, variations in amino acid sequences of XLG2 and the other two XLG proteins probably contribute to their different cofactor requirement.
In summary, our data demonstrate that all three XLG proteins exhibit both GTP binding and GTPase activity. Our results also revealed an interesting divergence from G␣ proteins because XLG proteins cannot utilize Mg 2ϩ as a cofactor. The unusual requirement for Ca 2ϩ in XLG protein GTP binding and hydrolysis raises the possibility that XLG proteins may be involved in the regulation of a variety of physiological and metabolic processes that are dependent on internal Ca 2ϩ con-centration, a well known component in many signal transduction chains in plants and other eukaryotes (68).
RTV1 Is an XLG2-interacting Protein-RTV1 was identified as an XLG2-interacting protein by a yeast two-hybrid screen using XLG2 as bait (Fig. 3A). We confirmed the interaction between RTV1 and XLG2 in vitro and in planta ( Fig. 3B and Fig.  4), indicating that RTV1 is a bona fide interacting protein of XLG2. Interestingly, the G␣-like C terminus of XLG2, rather than the N terminus, interacts with RTV1 (Fig. 3A). The interaction is not observed with the XLG2C(T475N) mutant (Fig. 3,  A and B), which lacks GTP binding activity (Fig. 1C). This suggests that GTP binding by XLG2 plays an important role in protein-protein interaction.
The RTV1 gene encodes a protein with a B3-like DNA binding domain, a domain that is unique to plants (56). The expression of RTV1 in vasculature is similar to that of other floral regulators, such as FT (69,70), establishing an overlapping expression profile between RTV1 and the RTV1 target, FT, and FIGURE 9. RTV1 has non-sequence-specific DNA binding activity in vitro that is enhanced by XLG2. A, EMSA of RTV1 using FLC, SOC1, FT, or VRN2 promoter regions. RTV1 protein (0 -57 nM; 0-2 g) and 400-bp radiolabeled double-stranded FLC, SOC1, FT, or VRN2 probe were incubated and then separated on a 5% Tris-borate EDTA-polyacrylamide gel. B, increased DNA binding by RTV1 in the presence of XLG2C but not in the presence of XLG2C(T475N). The reaction was performed as described in A together with 50 nM XLG2C or XLG2C(T475N) using FLC or SOC1 probes, respectively. C, the effect of GTP␥S binding by XLG2 in the presence of Ca 2ϩ . The reaction was performed as described in B together with 1 mM GTP␥S and 10 mM Ca 2ϩ using FLC or SOC1 probes, respectively.
suggesting that RTV1 is capable of regulating FT in locations where FT is expressed. The expression pattern of RTV1 and the nuclear localization of RTV1 are also very similar to those of XLG2 ( Fig. 5 and 6) (21), consistent with in planta interaction of the proteins.
RTV1 Promotes Floral Integrator Gene Expression and Flowering via an XLG2-related Mechanism-Overexpression of RTV1 led to early flowering (Fig. 7) through increased expression of floral integrator genes, such as FT and SOC1 (Fig. 7C). Consistent with this model, introgression of 35S::RTV1 into an ft background eliminated the early flowering phenotype (Fig. 8). Moreover, when RTV1 was overexpressed together with XLG2, the expression of the floral integrator genes was further enhanced, and these plants showed even earlier flowering compared with RTV1-only-overexpressing lines (Fig. 7). These results lead to a model whereby XLG2 interaction with RTV1 promotes RTV1 activity, leading to enhanced expression of the floral pathway integrator genes (Fig. 11).
XLG2 Promotes RTV1 Binding to Promoter Regions of Floral Integrator Genes-RTV1 exhibits strong DNA binding activities in vitro with DNA fragments derived from promoter regions of FLC, SOC1, FT, and VRN2 (Fig. 9) as well as with the promoter of the non-flowering related gene DRF (supplemental Fig. 6). Such non-sequence-specific binding to DNA fragments in vitro was also observed for VRN1, which also belongs to the B3 domain superfamily (58).
Swaminathan et al. (56) classified B3 genes into four families that are typified by ABI3 (abscisic acid-insensitive 3 (71)), ARF1 (auxin-responsive factor 1 (72)), RAV1 (related to ABI3/VP1 (73)), and REM (reproductive meristem). The 76 REM genes in Arabidopsis have been divided into six subgroups (A-F) based on phylogenetic analysis (56). RTV1 and VRN1 belong to the six genes of subgroup A of the REM family. VRN1 has been the only REM gene with known function (58). Unlike other B3 domain proteins, which show sequence-specific DNA binding (e.g. ABI3, RAV1, and auxin-response factors), both VRN1 and RTV1 bind to DNA in a non-sequence-specific manner in vitro (58). It is interesting to note that XLG2C enhances the DNA binding activity of RTV1 (Fig. 9, B and C). XLG2 in the presence of Ca 2ϩ and GTP␥S enhances the DNA binding activity of RTV1 even further (Fig. 9C), whereas the T475N mutant of XLG2C, which does not bind GTP, has no effect on the DNA binding activity of RTV1 (Fig. 9B). Taken together, these results suggest that XLG2 activation caused by GTP binding in the presence of Ca 2ϩ enhances the DNA binding activity of RTV1.
Although we did not observe sequence-specific DNA binding of RTV1 in vitro, specificity may be conferred by other components, such as other interacting proteins, present in planta but not in vitro. In fact, our ChIP analyses showed that RTV1 binds only to certain target genes in planta. In ChIP assays using 10-day-old 35S-FLAG-RTV1 (13) and 35S-FLAG-XLG2(1) 35S-RTV1-GFP seedlings, we consistently found that RTV1 directly interacts with FT and SOC1 promoters. Moreover, a slight increase of RTV1 at the FT promoter and an enrichment of RTV1 at the SOC1 promoter were observed in 35S-FLAG-XLG2(1) 35S-RTV1-GFP seedlings compared with 35S-FLAG-RTV1(13) (Fig. 10, B and C). This suggests that XLG2 facilitates RTV1 activation, which then results in transcriptional activation of FT and SOC1, an interpretation that is also consistent with our in vitro gel shift assays (Fig. 9C) and our gene expression assays (Fig. 7).
We did not observe any significant enrichment of RTV1 at FLC regions by ChIP assays (Fig. 10, B and C); nor did we observe any enhanced repression of FLC upon vernalization of 35S-FLAG-XLG2(1) 35S-RTV1-GFP lines (Fig. 7D). This suggests that RTV1 and XLG2 overexpression directly activate floral integrator genes in an FLC-independent manner. VRN1, which shares significant sequence homology with RTV1, also activates floral integrator genes via an FLC-independent pathway (58), and ectopic expression of RTV1 also accelerates flowering. These two proteins differ, however, in that VRN1, but not RTV1, has an additional function in stabilizing FLC repression postvernalization (58), and RTV1, but not VRN1, binds and is regulated by XLG2 (Fig. 11).
In summary, our work provides new insight into signaling by Ca 2ϩ and G proteins. We show that RTV1 and XLG2 function together in the regulation of flowering time. FT and SOC1 are direct targets of RTV1. It is likely that GTP-bound XLG2 and/or XLG2 cycling between GTP-and GDP-bound states, which occur in the presence of Ca 2ϩ , are required for enhanced RTV1 association with targets such as FT and SOC1. Circadian oscillations in intracellular free calcium ions ([Ca 2ϩ ] i ) occur and encode photoperiodic information (74,75). It is interesting to note that XLG2 and RTV1 transcripts exhibit in-phase circadian oscillation (see the Diurnal Search Tool Web site). It is tantalizing to speculate that XLG2-RTV1 mediates some of the outputs from the oscillations in [Ca 2ϩ ] i . However, the role of [Ca 2ϩ ] i oscillation in flowering time control is poorly understood. Further studies on the regulatory network of Ca 2ϩ , XLG2, and RTV1 should provide interesting insights into the roles of XLG proteins in Ca 2ϩ -mediated signaling pathways.