FTIP1 Is an Essential Regulator Required for Florigen Transport

FT-INTERACTING PROTEIN 1 is a novel protein that is involved in transporting florigen, a long-known mobile signal that induces flowering in plants in response to day length, from companion cells to sieve elements in the phloem of Arabidopsis.


Introduction
The transition to flowering, which is crucial for the reproductive success, is the most dramatic phase change in flowering plants.
Plants are able to adjust the timing of this transition in response to environmental conditions, such as photoperiod, temperature, and availability of nutrients. Classic experiments on the photoperiodic control of flowering in various plants have demonstrated that plant response to day length begins with the perception of photoperiod in leaves, followed by the transmission of a floral stimulus into the shoot apical meristem (SAM), where flowers are generated instead of leaves. Such mobile floral stimulus moving from leaves to the SAM was proposed as ''florigen'' in the 1930s [1]. Since then, tremendous efforts have been made to understand the molecular nature of this signal. Recent findings have suggested that the proteins encoded by FLOWERING LOCUS T (FT) in Arabidopsis and its orthologs in other plant species are part of the long-sought florigen [2][3][4][5][6].
FT encodes a member of the phosphatidylethanolaminebinding protein family and acts as a crucial regulator that relays flowering signals from the photoperiod pathway to floral meristem identity genes in Arabidopsis, which is a long-day (LD) facultative plant [7][8][9][10]. Under LDs, FT mRNA expression is activated by the CONSTANS (CO) transcriptional regulator in the vascular tissues of leaves and displays circadian rhythm [8,[11][12][13][14]. It has been suggested that long-distance movement of FT protein from leaves to the shoot apex through the phloem system plays a role in floral induction [2,4,5]. In the SAM, FT interacts with the bZIP transcription factor FLOWERING LOCUS D (FD), which in turn activates the downstream floral meristem identity genes such as APETALA1 (AP1) to initiate flower development [8,9]. Despite the remarkable progress in elucidating FT function, it is so far completely unknown whether and how FT protein transport is regulated. As the abundance of native FT protein is too low to be detectable, it has been hypothesized that simple diffusion of FT protein from companion cells to sieve elements might not be sufficient for transporting FT to the SAM [15].
Here we show that an endoplasmic reticulum (ER) membrane protein, FT-INTERACTING PROTEIN 1 (FTIP1), is required for FT protein transport in Arabidopsis. Loss of function of FTIP1 exhibits late flowering under LDs, which is partly due to the compromised FT movement to the SAM. FTIP1 and FT have similar mRNA expression patterns and subcellular localization, and they interact in vivo in phloem companion cells. Furthermore, FTIP1 is required for FT export from companion cells to sieve elements, thus affecting FT transport through the phloem to the SAM. Our results provide a mechanistic understanding of florigen transport and demonstrate that FT protein moves in a regulated manner and that FTIP1 is involved in mediating the export of FT protein from phloem companion cells to induce flowering.

FTIP1 Regulates Flowering Time under LDs
To understand how FT function is regulated, we performed yeast two-hybrid screening to identify proteins that interact with FT. Approximately 3 million yeast transformants were screened and 66 colonies were identified on the selective medium (Table  S1), among which a partial sequence belonging to an unknown protein with three C2 domains and one phosphoribosyltransferase C-terminal domain (PRT_C) was isolated ( Figure S1). The corresponding gene (At5g06850) was therefore named FT-INTERACTING PROTEIN 1 (FTIP1). We isolated two T-DNA insertional alleles, ftip1-1 (Salk_013179) and ftip1-2 (Salk_088086), from Arabidopsis Biological Resource Center ( Figure 1A). The fulllength FTIP1 transcript was undetectable in either homozygous mutant ( Figure 1B). Both ftip1-1 and ftip1-2 flowered late under LDs, but not under short days (SDs) ( Figure 1C,D; Table 1), suggesting that FTIP1 plays a role in mediating the effect of photoperiod on flowering. We transformed ftip1-1 with a genomic construct (gFTIP1) harboring a 5.1-kb FTIP1 genomic region including 2.1 kb of the upstream sequence, the 2.4-kb coding sequence, and 0.6 kb of the downstream sequence ( Figure S2A). Most ftip1-1 gFTIP1 T1 transformants exhibited similar or slightly late flowering time as compared to wild-type plants ( Figure 1E), demonstrating that FTIP1 is responsible for promoting flowering particularly under LDs.

Gene Expression and Subcellular Localization of FTIP1
We tested FTIP1 expression in various tissues of wild-type plants using quantitative real-time PCR and found its highest expression in leaves and stems ( Figure S3). To examine the detailed expression pattern of FTIP1, we generated a FTIP1:b-glucuronidase (GUS) reporter construct in which the same 2.1-kb FTIP1 upstream sequence included in gFTIP1 for the gene complementation test was fused to the GUS reporter gene ( Figure S2A). We created 23 independent FTIP1:GUS lines, most of which showed similar GUS staining patterns. A representative line was selected to monitor the detailed expression pattern of FTIP1. FTIP1:GUS showed specific GUS staining in vascular tissues of various plant organs ( Figure S2B-H). Notably, in developing seedlings during the floral transition occurring 7 d after germination, FTIP1:GUS and FT:GUS [14] shared similar GUS staining patterns in vascular tissues of cotyledons and rosette leaves, although the former had a relatively broad and intensive staining pattern (Figure 2A). A crosssection of a primary leaf vein revealed that FTIP1:GUS expression was specifically located in the phloem including companion cells ( Figure 2B), which is similar to the FT:GUS expression pattern [14]. Neither FTIP1:GUS nor FT:GUS was expressed in the SAM ( Figure 2C,D; Figure S2C) [14]. Furthermore, the late-flowering phenotype of ftip1-1 was rescued by the expression of FTIP1 coding sequence driven by the promoter of SUCROSE TRANS-PORTER 2 (SUC2) ( Figure 3A), which is active specifically in phloem companion cells [16]. These results suggest that FTIP1 functions in the phloem to promote flowering.
Given that FTIP1 functions in flowering time control, we investigated whether its expression is regulated by known flowering genetic pathways. FTIP1 expression was not regulated by photoperiod and did not exhibit an obvious circadian rhythm under LDs ( Figure S4A,E). Similarly, vernalization treatment did not affect FTIP1 expression ( Figure S4B), and GA treatment did not affect FTIP1 expression and the flowering phenotype of ftip1-1 ( Figure S4C,D). In addition, FTIP1 expression was also not altered in several mutants tested in known flowering genetic pathways ( Figure S5). These observations imply that flowering signals may not regulate FTIP1 function through affecting its mRNA levels.
Next, we examined the subcellular localization of FTIP1 through monitoring the signal of the green fluorescent protein (GFP) fused with FTIP1 under the control of FTIP1 or SUC2 promoter, respectively. Both constructs could rescue the late flowering phenotype of ftip1-1 ( Figure 3A). However, we could not detect fluorescent signal from either SUC2:FTIP1:GFP ftip1-1 or FTIP1:FTIP1:GFP ftip1-1 transgenic lines, indicating that FTIP1 protein might be present at very low abundance in plant cells. Alternatively, we transiently expressed 35S:FTIP1:GFP with various fluorescent protein-tagged organelle markers in N. benthamiana leaf epidermal cells and found that FTIP1:GFP was mostly colocalized with an endoplasmic reticulum (ER) marker (Figure 3B,C; Figure S6) [17]. We did not observe FTIP1:GFP signals in the nucleus ( Figure 3C). Notably, at the cell wall, FTIP1:GFP colocalized with callose deposition stained with aniline blue, which marks the position of plasmodesmata ( Figure 3C).
To precisely localize FTIP1, we performed immunoelectron microscopy on an FTIP1:4HA:FTIP1 ftip1-1 transgenic line, in which FTIP1:4HA:FTIP1 was able to rescue the flowering defect of ftip1-1 ( Figure 3A). The result revealed that 4HA:FTIP1 was specifically localized in phloem companion cells ( Figure 3D) and plasmodesmata between companion cells and sieve elements ( Figure 3E,F; Figure S7), where the ER membrane runs through.

FTIP1 Interacts with FT in Phloem Companion Cells
Several pieces of evidence, including the initial identification of FTIP1 as an FT interacting partner, similar tissue expression pattern of FTIP1 and FT, and similar late-flowering phenotype exhibited by ftip1 and ft mutants specifically under long days, point to a possible role of FTIP1 in mediating FT function in the control of flowering time. Thus, we further carried out a detailed analysis of the interaction between FTIP1 and FT. As revealed in our yeast two-hybrid screening, a truncated FTIP1 protein devoid of the

Author Summary
The transition to flowering is the most dramatic phase change in flowering plants and is crucial for reproductive success. Such a transition from vegetative to reproductive growth is controlled by seasonal changes in day length. Studies originally performed in the 1930s were the first to suggest that day length is perceived by a plant's leaves; by contrast, flower formation takes place in the shoot apical meristem (the tip of the shoot that gives rise to plant organs, such as leaves and flowers). The term ''florigen'' was later proposed to describe a mobile floral stimulus that moves from leaves to the shoot apical meristem to induce flowering. It is only recently that FLOWERING LOCUS T (FT) in Arabidopsis, and its orthologs in various other plant species, was identified as being florigen, but how florigen is transported in plants remains completely unknown. Here, we report that a novel ER membrane protein, FT-INTERACTING PROTEIN 1 (FTIP1), interacts with FT in companion cells of the phloem (a specialized type of parenchyma cell in the phloem of the plant's vascular system) and mediates FT protein movement from companion cells to sieve elements (the conducting cells of the phloem), thus affecting FT transport to the shoot apical meristem in Arabidopsis. To our knowledge, this study reveals the first regulator that is required for florigen transport and offers new insights into possible florigen transport mechanisms in other flowering plants.
PRT_C domain interacted with FT in both yeast two-hybrid and GST pull-down assays ( Figure 4A-C), whereas no interaction was detected using the full-length FTIP1 (unpublished data). Since the PRT_C domain of FTIP1 was predicted to be a membranetargeted domain according to a protein topology analysis ( Figure  S1), the full-length FTIP1 protein might not be in the membranebound state in yeast cells or under in vitro conditions and thus might undergo inappropriate folding, which prevents its interaction with FT. Alternatively, in yeast two-hybrid assay the fulllength FTIP1 protein might be membrane-bound and unable to reconstitute a functional transcription factor in the yeast nucleus to drive the reporter gene expression.
We transiently expressed 35S:FTIP1:GFP with 35S:FT:RFP in N. benthamiana leaf epidermal cells and revealed that both FTIP1:GFP and FT:RFP were colocalized to ER connected to the nuclear envelope ( Figure S8). However, in contrast to FTIP1:GFP, FT:RFP was also localized in the nucleus, which is consistent with a previous observation [9]. These results indicate that FTIP1 may not directly mediate FT function in transcriptional regulation of other target genes.
To test whether and how FT interacts with FTIP1 in vivo, we performed in situ Proximity Ligation Assay (PLA) [18], in which dual recognition of target proteins by pairs of affinity probes generates an amplifiable DNA reporter molecule that serves as a surrogate marker for interacting proteins, to examine the subcellular localization of FT and FTIP1 interaction at single-

FTIP1 Controls the Export of FT Protein from Phloem Companion Cells to Sieve Elements
The findings on the interaction between FT and FTIP1, and FTIP1 localization to ER and plasmodesmata prompted us to hypothesize that FTIP1 may regulate FT export from phloem companion cells. To this end, we first examined whether FTIP1 affects FT transport to the SAM during the floral transition. We generated a SUC2:FT:GFP transgenic line as previously described [2]. As this transgenic allele could significantly rescue the lateflowering phenotype of the FT null mutant, ft-10 (Table 1), we further crossed this SUC2:FT:GFP allele with ftip1-1 and 35S:FTIP1. Confocal analysis of the distribution of FT:GFP fusion protein revealed that in 11-d-old seedlings, which were undergoing the floral transition, FT:GFP was clearly detected in the inner cone-like region of the SAM in wild-type background, but not in ftip1-1 ( Figure 5A). In contrast, the distribution of free GFP protein  was comparably undetectable in the inner region of the SAM in wild-type and ftip1-1 ( Figure S9A), indicating a specific effect of FTIP1 on FT:GFP distribution in the SAM during the floral transition. In agreement with the above observations, SUC2: FT:GFP ftip1-1 flowered later than SUC2:FT:GFP (Table 1). Since the abundance of FT:GFP mRNA and protein in SUC2:FT:GFP was not downregulated in ftip1-1 ( Figure 6A-C), the difference in FT:GFP distribution in the SAM between wild-type and ftip1-1 plants suggests a role of FTIP1 in regulating FT transport rather than FT mRNA or protein abundance.
As FTIP1 was expressed in the phloem ( Figure 2) and its protein was localized in phloem companion cells ( Figure 3D-F; Figure 4D), we examined whether FTIP1 affects FT transport from companion cells to sieve elements in a newly created SUC2:FT:9myc line in wild-type and ftip1-1 backgrounds using immunoelectron microscopy ( Figure 5B). This SUC2:FT:9myc transgenic allele substantially rescued the late-flowering phenotype of ft-10 ( Table 1), indicating that FT:9myc retains the biological function of endogenous FT protein. Signals corresponding to FT:9myc could be specifically detected by anti-myc antibody in the phloem of the transgenic plants harboring SUC2:FT:9myc ( Figure 5B; Figure  S10). Quantitative analysis of labeling density of FT:9myc in companion cell-sieve element complexes showed that although all sections from SUC2:FT:9myc and SUC2:FT:9myc ftip1-1 displayed    FT:9myc labeling in companion cells ( Figure 5B, lower right panel), there was an approximate 3-fold enrichment of labeling density in ftip1-1 over wild-type background ( Figure 5B, lower left panel). More importantly, we detected FT:9myc labeling in sieve elements in nearly 80% of the wild-type sections, whereas only 4% of ftip1-1 sections displayed FT:9myc labeling in sieve elements ( Figure 5B, lower right panel). In addition, the labeling density of FT:9myc in sieve elements was much higher in wild-type than in ftip1-1 ( Figure 5B, lower left panel). Thus, in the absence of FTIP1, FT:9myc accumulated in companion cells and its transport to sieve elements was compromised. In agreement with this result, SUC2:FT:9myc ftip1-1 displayed later flowering than SUC2:FT:9myc ( Table 1). As ftip1-1 also delayed flowering in SUC2:FT and SUC2:GFP:CO where CO directly promotes the endogenous FT expression (Table 1) [12], it seems that FTIP1 similarly affects the promotive effect of untagged FT protein on flowering as other FT fusion proteins used in this study. These observations support that FTIP1 regulates FT export from phloem companion cells to sieve elements, thus affecting FT transport through the phloem to the SAM. Consistent with this conclusion, the early-flowering phenotype caused by expression of FT or FT:GFP under the control of the KNAT1 promoter, which is active in the SAM [12], was not affected by ftip1-1 (Table 1). Unlike other flowering promoters, overexpression of FTIP1 surprisingly caused late flowering ( Figure S11; Table 1). Confocal analysis showed that the expression of FT:GFP protein in the inner region of the SAMs of 11-d-old seedlings was substantially lower in 35S:FTIP1 than in wild-type plants ( Figure 5A; Figure S9A). In the number of sections examined for each genotype is listed above the histogram. The right panel shows the average number of red spots per phloem companion cell. Statistical analysis was performed using a two-tailed unpaired Student's t test. The results are considered statistically significant at p,0.05. doi:10.1371/journal.pbio.1001313.g004 primary leaf vein, FT:GFP driven by the SUC2 promoter was exclusively detected in phloem companion cells in wild-type background, whereas in 35S:FTIP1, the distribution of FT:GFP signals was detected in both phloem companion cells and xylem parenchyma cells ( Figure 5C). However, the free GFP driven by the SUC2 promoter remained in phloem companion cells of 35S:FTIP1 as compared to wild-type plants ( Figure S9B). These observations demonstrate that that ectopic expression of FTIP1 specifically deregulates the transport of FT:GFP protein out of the phloem system, an effect previously shown for a viral movement protein [19]. This could compromise the eventual distribution of FT:GFP in the SAM of 35S:FTIP1 and thus delay flowering.  [8,9,20]. As expected, the expression of these genes was downregulated in ftip1-1 in which FT transport is defective ( Figure  S12A). Surprisingly, FT expression was also downregulated in ftip1-1, whereas the expression of CO, a direct upstream activator of FT, was not significantly changed ( Figure S12). As FTIP1 protein is not localized in the nucleus, it is unlikely that FTIP1 directly controls FT transcription. To address whether FTIP1 could regulate the stability of FT transcripts, we compared the levels of FT transcripts generated from the native and SUC2:FT:GFP transgenic loci. Although steady-state levels of native FT expression were downregulated in ftip1-1, total FT expression including native FT and transgenic FT:GFP expression remained unchanged in SUC2:FT:GFP ftip1-1 ( Figure 6A,B). In addition, the abundance of FT:GFP fusion protein remained unchanged in wild-type and ftip1-1 backgrounds ( Figure 6C). These results suggest that FTIP1 may not be directly involved in regulating FT mRNA or protein stability. Meanwhile, we observed downregulation of native FT expression in SUC2:FT:GFP ( Figure 6A,B) and reduced FT:GUS staining in SUC2:FT ( Figure 6D). These results are in agreement with the observation in a previous study [2] implying that an excessive accumulation of FT protein in phloem companion cells caused by the SUC2 promoter might directly or indirectly result in a reduction in native FT mRNA expression through a negative feedback loop. This may explain the observed downregulation of native FT expression in ftip1-1, where defective export of FT protein causes accumulation of FT protein in phloem companion cells ( Figure 5B).

Discussion
Our results have demonstrated that FTIP1 and FT share similar mRNA expression patterns and subcellular localization, and that they interact in vivo in phloem companion cells. During the floral transition, the FT:GFP accumulation at the SAM is compromised in ftip1 mutants, which eventually exhibit late flowering under LDs. Consistently, FTIP1 is required for FT:9myc export from phloem companion cells to sieve elements, thus affecting the flowering time of SUC2:FT:9myc. In addition, overexpression of FTIP1 causes the transport of FT:GFP out of the phloem system, which also results in late flowering. These observations suggest that FT protein moves from phloem companion cells to sieve elements in a regulated manner and that a subtle regulation of FTIP1 activity is indispensable for the export of FT protein from phloem companion cells to induce flowering.
We envisage that in addition to FTIP1 and FT, florigen transport should involve other relevant regulators. First, although the transport of FT:9myc protein from phloem companion cells to sieve elements in ftip1-1 is significantly compromised, it is not completely abolished. This implies either that there is a basal level of diffusion of FT protein or that FT transport also depends on other regulators that share a redundant function with FTIP1 in mediating FT export from phloem companion cells. Furthermore, previous examinations of the spatial distribution of FT:GFP fusion protein in both Arabidopsis and rice have shown that FT:GFP accumulates in the rib zone beneath the SAM in a conical shape [2,3], indicating that the movement of FT protein from phloem to the rib zone is not a simple diffusion process. As FTIP1 is clearly not expressed in the whole SAM ( Figure 2C,D), regulation of FT protein transport from the phloem stream to the rib zone might also involve other regulators. The requirement of other regulators for FT protein transport is supported by the genetic analysis showing that ft mutants display much later flowering than ftip1 (Table 1). Potential candidates for these regulators include FTIP1 homologs ( Figure S13A) because some combinations of ftip1 with loss-of-function mutants of FTIP1 homologs show much later flowering than any single mutant (unpublished data).
Second, the late-flowering phenotype of ft mutants is further enhanced by ftip1-1 ( Table 1), indicating that FTIP1 may be required for transporting other flowering molecules in addition to FT. A potential candidate could be TWIN SISTER OF FT (TSF), which encodes another phosphatidylethanolamine-binding protein with very high sequence similarity with FT [21,22]. Mutation of TSF further enhances the late flowering of ft mutants, and the resulting double mutants fail to accelerate flowering in response to LD conditions [21,22]. The expression domain of TSF also overlaps with that of FTIP1 [21]. Furthermore, loss of function of FTIP1 does not further delay flowering of ft-10 tsf-1 under LD conditions (Table 1). These data support that TSF functions redundantly with FT and could be another molecule whose transport is affected by FTIP1.
As both FTIP1 and FT proteins are localized to ER, regulation of FT movement by FTIP1 across the border between companion cells and sieve elements might be partly mediated by a continuous ER network within plasmodesmata [23,24]. In plasmodesmata, the ER becomes appressed to form the central axial desmotubule surrounded by the plasma membrane continuum between adjacent cells [25]. Although it has been suggested that the desmotubule is not the main route for plasmodesmatal transport, some molecules are known to be transported through this channel [26]. In contrast, the space between the desmotubule and the plasma membrane, which is referred as the cytoplasmic sleeve, is the proposed place where the general trafficking of molecules and ions occurs [25]. Because FTIP1 possesses a membrane-targeted PRT_C domain ( Figure S1) and is localized to plasmodesmata ( Figure 3C,E,F), it is likely that the C-terminus of FTIP1 is anchored to the desmotubule. How FTIP1 is oriented in plasmodesmata is an important question as its N-terminus, which is included in the region that interacts with FT protein ( Figure 4A-C), might face either the cytoplasmic sleeve or the interior of the desmotubule. Further addressing this question will help to identify the route where FT protein passes through plasmodesmata and other possible factors involved in FT transport. Based on the size of FT:GFP, the route through the cytoplasmic sleeve might be possible for FT transport as the current understanding is that molecules larger than 27 kDa do not move readily through desmotubule [23].
The presence of C2 domains and a transmembrane domain in FTIP1 and its close homologs in Arabidopsis makes them topologically resemble synaptotagmins ( Figure S13A) that constitute a family of membrane-trafficking proteins widely found in plants and animals. In Arabidopsis, the synaptotagmin SYTA has been shown to regulate endosome recycling and movement protein-mediated trafficking of plant virus genomes through plasmodesmata [27]. Our finding on the function of FTIP1 in mediating FT export from phloem companion cells to sieve elements, together with the proposed SYTA function, implies that synaptotagmin-like proteins may serve as essential regulators that mediate the transport of macromolecules in plants. Another FTIP1-like gene, QUIRKY (QKY; At1g74720), has been suggested to contribute to plant organ organogenesis mediated by the receptor-like kinase STRUBBELIG [28], implying a role of QKY in intercellular signaling. As FTIP1-like proteins are highly conserved in the angiosperms (Figure S13B), further investigation of FTIP1 and its homologs might shed light on the conserved mechanisms underlying which flowering plants regulate cell-to-cell communication to coordinate the growth and development.

Yeast Two-Hybrid Assay
All vectors used in yeast two-hybrid assays were from Clontech. The coding sequence of FT was cloned into pGBKT7 to produce BD-FT, which was used as a bait to screen cDNA library (CD4-30 from ABRC) for identifying interacting proteins of FT. Selection was performed on medium lacking histidine, tryptophan, and leucine (SD-His/-Trp/-Leu) supplemented with 30 mM 3-amino-1, 2, 4-triazole. To verify the interaction between FT and FTIP1, various versions of FTIP1 coding sequences were cloned into pGADT7. The resulting vectors were co-transformed with BD-FT, and the transformed cells were selected on SD-His/-Trp/-Leu medium supplemented with 30 mM 3-amino-1, 2, 4-triazole. bgalactosidase assays were performed according to the Yeast Protocols Handbook (Clontech).

GST In Vitro Pull-Down Assay
The cDNA encoding FT was cloned into the pGEX-4T-1 vector (Pharmacia) and introduced into E. coli Rosetta (DE3) (Novagen). Transformed cells were cultured until the OD 600 nm reached 0.6, and IPTG was added to a final concentration of 0.6 mM to start induction. After overnight induction at 16uC, cells were collected and homogenized with lysis buffer (10 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% Triton-100, and 10 mM PMSF). The soluble GST fusion proteins were extracted and immobilized on glutathione sepharose beads (Amersham Biosciences) for subsequent pull-down assays. The FTIP1 Nterminal fragment containing the three C2 domains (N501) was cloned into pGBKT7 vector (Clontech). The resulting plasmid was added to the TNT T7 Quick Coupled Transcription/Translation Systems (Promega) to synthesize myc-FTIP1(N501) protein. The resulting fusion protein was then incubated with the immobilized GST and GST fusion proteins. Proteins retained on the beads were resolved by SDS-PAGE and detected with anti-myc antibody (Santa Cruz Biotechnology).

Transient Expression of Proteins in Nicotiana Benthamiana Leaf Epidermal Cells
The overnight Agrobacterium cultures with a desired expression vector (35S:FTIP1:GFP, various RFP-or CFP-tagged organelle markers, 35S:FT:RFP, or 35S:GFP) were harvested and resuspended with infiltration buffer (10 mM MES pH 5.6, 10 mM MgCl 2 , and 100 mM acetosyringone) with OD 600 nm at 0.4. To compare protein localization, equal volumes of infiltration solutions containing different expression vectors were mixed together and incubated for 3 h at room temperature. Infiltration solutions were infiltrated into the abaxial surface of 3-wk-old tobacco (Nicotiana benthamiana) leaves with syringes. The leaves were examined 2 d after infiltration under a confocal microscope.

GUS Staining
Tissues were infiltrated with staining solution (50 mM sodium phosphate buffer, pH 7.0, 0.5 mM potassium ferrocyanide, 0.5 mM potassium ferricyanide, and 0.5 mg/mL X-Gluc) in a vacuum chamber, and subsequently incubated with staining solution at 37uC overnight. For sectioning, samples were dehydrated through an ethanol series, an ethanol/histoclear series, and finally embedded in paraplast (Sigma). Samples were then orientated and sectioned at a thickness of 3 mm with a microtome.

Expression Analysis
Total RNA was isolated with RNeasy Plant Mini Kit (Qiagen) and reverse-transcribed with ThermoScript RT-PCR System (Invitrogen) according to the manufacturers' instructions. Realtime PCR was performed in triplicates on 7900HT Fast Real-Time PCR system (Applied Biosystems) with SYBR Green PCR Master Mix (Applied Biosystems). The difference between the cycle threshold (Ct) of the target gene and the Ct of TUB2 (DCt = Ct target gene 2Ct tubulin ) was used to obtain the normalized expression of target genes, which corresponded to 2 2DCt . Expression analysis was performed with at least three biological replicates. Primers for real-time PCR are listed in Table S2.

In Vivo Protein Interaction Assay
Plant tissues were collected and fixed with ice-cold 4% paraformaldehyde (PFA; Sigma-Aldrich) at pH 7.0 in a vacuum chamber. A serial PFA/sucrose change was applied till the tissues were finally equilibrated in PFA with 20% sucrose. Tissues were then embedded in 1.5% agarose gel, placed onto the microtome tissue holder, and flash frozen with liquid nitrogen. Tissues were cut in cryo-microtome with 20 mm thickness, and the sections were placed on slides. After complete drying, the slides were rehydrated with 100 mM Tris pH 8, 50 mM EDTA, and permeabilized with proteinase K (1 mg/ml) in the same buffer for 10 min at room temperature. Slides were washed with 2 mg/ml glycine followed by washing with PBS solution. Chlorophyll molecules were subsequently removed by incubating the slides with 1:1 acetone/ methanol mixture twice for 5 min. After drying, slides were rehydrated with PBS and finally treated with 4% paraformaldehyde for 10 min followed by washing with PBS solution.
In situ Proximity Ligation Assay (PLA) was performed with Duolink kit (Olink Bioscience) with minor modifications. The above treated slides were firstly blocked with 2% Bovine Serum Albumin in 100 mM Tris pH 7.5, 150 mM NaCl, and 0.3% Triton X-100 for 45 min at 37uC, and probed with the mixture of anti-GFP and anti-HA antibodies diluted in the blocking solution (1:60) for 45 min at 37uC. The slides were washed three times and probed with diluted PLUS and MINUS PLA probes for 1 h at 37uC and subsequently washed 5 times. The slides were further incubated with the ligation solution, washed, and subsequently incubated with the amplification-polymerase solution with all components provided in the kit. After signal amplification, the slides were washed and mounted with PBS solution for further observation.

Immunogold Transmission Electron Microscopy
Samples were fixed with paraformaldehyde-glutaraldehyde solution (2% and 2.5%, respectively) and imbedded with LR white medium (EMS). Ultra-thin sections (85 nm) were cut and mounted on nickel grids. The grids were blocked with 1% BSA in TTBS (20 mM Tris, 500 mM NaCl, and 0.05% Tween-20, pH 7.5) for 30 min and subsequently incubated with anti-HA or anti-myc antibody at 1:5 (v/v) for 1 h at room temperature. The grids were washed with TTBS for three times and further incubated with 15 nm gold-conjugated goat anti-mouse antibody (EMS) that was diluted 1:20 with blocking solution. After 40 min of incubation, the grids were washed with TTBS for three times and with distilled water twice. Tissue staining was performed with 2% uranyl acetate for 15 min at room temperature, and pictures were taken by transmission electron microscope (Jeol JEM-1230).
For quantitative analysis of immunogold labeling, micrographs of randomly photographed immunogold-labeled transverse sections of the first rosette leaves of 15-d-old seedlings with various genetic backgrounds were measured as previously reported [32]. The data were presented as the mean number of gold particles per mm 2 plus or minus standard deviation. The projected cell area was measured by a LI-3100C area meter (Li-Cor). We analyzed 56 individual sections from eight different leaves of each genotype for calculating the density of gold particles over the projected cell area. Statistical analysis was performed using a two-tailed unpaired Student's t test. Two-tailed test results were considered statistically significant at p,0.05.