Combined Bimolecular Fluorescence Complementation and Fo¨rster Resonance Energy Transfer Reveals Ternary SNARE Complex Formation in Living Plant Cells

Various ﬂuorophore-based microscopic methods, comprising Fo¨rster resonance energy transfer (FRET) and bimolecular ﬂuorescence complementation (BiFC), are suitable to study pairwise interactions of proteins in living cells. The analysis of interactions between more than two protein partners using these methods, however, remains difﬁcult. In this study, we report the successful application of combined BiFC-FRET-ﬂuorescence lifetime imaging microscopy and BiFC-FRET-acceptor photobleaching measurements to visualize the formation of ternary soluble N -ethylmaleimide-sensitive factor attachment receptor complexes in leaf epidermal cells. This method expands the repertoire of techniques to study protein-protein interactions in living plant cells by a procedure capable of visualizing simultaneously interactions between three ﬂuorophore- tagged polypeptide partners.

Many biological processes rely on the dynamic assembly and disassembly of multicomponent protein complexes. Experimental methods based on genetically encoded fluorophores are widely used to study the subcellular localization and binary protein-protein interactions in living cells. The subcellular localization of proteins can be visualized by translationally fusing them to fluorophores, which are now available across a wide spectral range (Shaner et al., 2005). Fluorescence microscopy allows imaging and tracing of these proteins in real time in living cells. However, the resolution of a conventional epifluorescence microscope or a confocal laser scanning microscope is insufficient to resolve distances smaller than approximately 200 nm between biological macromolecules and to demonstrate their vicinity at the molecular scale (Bhat et al., 2006;Held et al., 2008).
Two fluorophore-based methods are commonly employed to study protein-protein interactions on the stage of a light microscope and overcome its limita-tions: bimolecular fluorescence complementation (BiFC; Hu et al., 2002;Bracha-Drori et al., 2004;Walter et al., 2004) and Fö rster resonance energy transfer (FRET; Hink et al., 2002;Chen et al., 2003;Russinova et al., 2004). The BiFC approach is based on the restoration of an intact fluorophore from its nonfluorescent N-and C-terminal domains. When the two complementing fragments are fused to potentially interacting proteins, close proximity of the proteins can bring the N-and C-terminal fluorophore domains into a favorable position and orientation and thereby facilitate their association into a functional fluorophore (Hu et al., 2002;Kerppola, 2008). Since only standard fluorescence microscopic equipment is required for this technique, it is a popular method to analyze protein-protein interactions (Weinthal and Tzfira, 2009).
An alternative procedure exploits a physical phenomenon referred to as FRET to obtain information about the molecular vicinity of fluorophore-labeled proteins. FRET occurs when a donor fluorophore is brought into close proximity (less than 10 nm) of a suitable acceptor fluorophore. In this case, the donor can transmit its excitation energy to the acceptor. The extent of FRET depends on various parameters, such as the distance between donor and acceptor, their spectral properties, the relative orientation of the donor and acceptor transition dipoles, and the refractive index of the medium (for a more detailed introduction to the theory of FRET, see "Materials and Methods"; Clegg, 1996;Periasamy and Day, 2005;Biskup et al., 2007). The extent of FRET can be estimated by several means, such as by assessing the sensitized emission (Shah et al., 2001(Shah et al., , 2002, by determining the donor fluorescence before and after acceptor photobleaching (APB; Bhat et al., 2005), or by measuring the fluores-cence lifetime of the donor (Russinova et al., 2004;Bhat et al., 2005;Tonaco et al., 2006;Shen et al., 2007;Osterrieder et al., 2009). Depending on the method, FRET measurements require more or less sophisticated microscopic equipment.
Many biological processes involve protein complexes composed of more than two proteins, which are difficult to study with the experimental approaches outlined above. For example, biologically active N-ethylmaleimide-sensitive factor attachment receptor (SNARE) complexes are composed of three distinct protein partners. SNAREs, together with accessory proteins and cytosolic calcium, catalyze membrane fusion events in eukaryotic cells. A binary or ternary t-SNARE complex composed of a Qa-SNARE, also called syntaxin, and Qb-and Qc-SNARE domains present in either one protein (synaptosome-associated protein ) or as two separate units is located at the target membrane. This t-SNARE complex interacts with an R-SNARE (vesicle-associated membrane protein [VAMP]) present on the vesicle, and the formation of the fusogenic SNARE complex brings the opposing membranes into close proximity (Jahn and Scheller, 2006;Leabu, 2006). Besides their role in cellular homoeostasis, SNAREs play a major role in various plant biological processes, such as cytokinesis, gravitropism, and defense against pathogens (Lipka et al., 2007).
The orthologous Arabidopsis (Arabidopsis thaliana) and barley (Hordeum vulgare) syntaxins, AtPEN1 and HvROR2, respectively, are essential in restricting the invasion of nonadapted powdery mildew species or in broad-spectrum powdery mildew resistance conferred by loss-of-function mlo mutants, respectively (Collins et al., 2003;Assaad et al., 2004). Both syntaxins form a binary t-SNARE complex with the orthologous SNAP-25-like proteins AtSNAP33 and HvSNAP34 (Collins et al., 2003;Kwon et al., 2008), which in turn interact with the R-SNAREs AtVAMP721/AtVAMP722 and HvVAMP721, respectively, by adopting an authentic ternary SNARE complex . In addition to a range of other polypeptides, the mentioned Arabidopsis and barley SNARE partners focally accumulate below the site of attempted fungal penetration (Assaad et al., 2004;Bhat et al., 2005;Kwon et al., 2008), which probably precedes their extracellular deposition in the paramural space (Meyer et al., 2009).
In this study, we combined BiFC-FRET measurements based on cerulean fluorescent protein (CrFP; Rizzo et al., 2004) as a donor and reconstituted yellow fluorescent protein (YFP; Walter et al., 2004;Schü tze et al., 2009) as an acceptor to study the formation of a ternary protein assembly. We concentrated on the wellcharacterized SNARE complexes described above that play a decisive role in antifungal defense. FRET was detected by measuring the fluorescence lifetime of the donor in each pixel of the sample in a fluorescence lifetime imaging (FLIM) setup. Occurrence of FRET was corroborated by APB experiments. Our data demonstrate the technical feasibility of combined BiFC-FRET-FLIM and BiFC-FRET-APB assays to study ternary protein complexes in living plant cells.

HvROR2 Is Present in Large SDS-Resistant, Heat-Sensitive Protein Complexes
Binary interactions between HvROR2, HvSNAP34, and HvVAMP721 have been demonstrated previously (Collins et al., 2003;Kwon et al., 2008). Before assessing ternary SNARE complex formation by BiFC and FRET, we aimed to demonstrate HvROR2-containing ternary complexes in barley. In analogy to experiments performed by Kwon et al. (2008), we detected the presence of HvROR2-containing SDS-resistant, heat-sensitive ternary SNARE complexes by biochemical means in 7-d-old barley leaves (Fig. 1). Thus, like its ortholog AtPEN1, HvROR2 forms ternary SNARE complexes in vivo.

Colocalization of Fluorophore-Tagged SNARE Fusion Proteins in Epidermal Cells
We studied the subcellular localization of the barley SNAREs HvSNAP34, HvROR2, and HvVAMP721 by expression of fluorophore-tagged versions of these proteins in barley leaf epidermal cells. Owing to the longevity of detached barley leaves used for transformation, lower autofluorescence levels, and higher expression levels, we performed most experiments of this study in this species and validated essential observations in Arabidopsis. The translational fusion variants of CrFP (Rizzo et al., 2004) to the N termini of HvROR2 (CrFP-HvROR2) or HvVAMP721 (CrFP-HvVAMP721) predominantly localized to the cell pe- Figure 1. The HvROR2 protein is present in heat-sensitive, but SDSresistant, complexes in barley leaves. Total protein extracts were prepared from barley wild-type leaves. Following SDS-PAGE (50 mg lane 21 ) on either 10% or 7.5% gels, HvROR2 was detected by immunoblot using an anti-HvROR2 antiserum (Collins et al., 2003). The 10% gel allows visualization of the HvROR2 monomer, while the 7.5% gel illustrates the high-M r HvROR2-containing complexes. Note the absence of these complexes upon boiling of the sample prior to gel loading. riphery, most likely the plasma membrane (PM; Fig. 2, A and B). Aside from the PM, CrFP-HvROR2 and in particular CrFP-HvVAMP721 were often visible in mobile spots close to the PM (Supplemental Fig. S1, A and B). The C-terminal fusion of HvSNAP34 with CrFP (HvSNAP34-CrFP) was predominantly present at the cell periphery but also visible in cytoplasmic areas (Fig. 2C). We observed similar subcellular localizations for the Arabidopsis SNAREs AtPEN1, AtSNAP33, and AtVAMP722 upon coexpression of fluorophore-tagged versions of these proteins in epidermal cells of Arabidopsis rosette leaves (compare Fig. 2, A-C, with Supplemental Fig. S4, and Supplemental Fig. S1,A and B,with Supplemental Fig. S1,C and D).

Analysis of Pairwise Interaction of SNARE Proteins Using the BiFC Assay
Pairwise interactions between either the abovementioned barley SNAREs or their Arabidopsis orthologs were previously revealed in yeast two-hybrid studies and in planta by FRET-FLIM analysis and biochemical methods (Collins et al., 2003;Kwon et al., 2008;Pajonk et al., 2008). With the ultimate goal to visualize ternary protein-protein interactions, we first aimed at corroborating these binary interactions using the BiFC assay (Hu et al., 2002;Bracha-Drori et al., 2004;Walter et al., 2004). Fusions of interacting proteins with split YFP halves that are able to form a functional fluorophore can then serve as an acceptor in FRET experiments involving a third interaction partner, which is fused to CrFP as donor fluorophore. Translational fu-sions of the SNARE partners to nonfluorescent N-and C-terminal halves of YFP were coexpressed in barley leaf epidermal cells, and restoration of YFP fluorescence was assessed. We tested all possible pairwise combinations of translational fusions of the split YFP halves to the SNARE proteins by epifluorescence microscopybased detection of the YFP signal (Supplemental Fig. S2). We noted that only a few tested combinations (e.g. a fusion of the N-terminal half of YFP [ N YFP] to the N terminus of HvROR2 [ N YFP-HvROR2] coexpressed with a fusion of the C-terminal half of YFP [ C YFP] to the C terminus of HvSNAP34 [HvSNAP34-C YFP]) provided bright BiFC fluorescence, suggesting that in these fusion protein pair constellations the fluorophore halves are in a sterical orientation that is favorable for complementation of the YFP fluorophore. In the subsequent analysis, we focused on these fusion protein combinations, because only intact functional YFP fluorophores can act as an acceptor in FRET experiments.
Examination by confocal laser scanning microscopy of the BiFC couples with the brightest YFP fluorescence ( Fig. 2, D-F) revealed that the interacting proteins were not as evenly distributed at the PM as in the case of the single proteins but were often present in large clusters of vesicle-like compartments (VLCs; arrows in Fig. 2, D-F). Coexpression of N YFP-HvROR2 together with HvSNAP34-C YFP and N YFP-HvVAMP721 with HvSNAP34-C YFP resulted in VLCs that were larger and showed more intense fluorescence than in case of N YFP-HvROR2 coexpressed with C YFP-HvVAMP721 (Fig. 2, compare D with E and F). Although weaker than in the VLCs, YFP fluorescence could still be observed at the cell periphery (Supplemental Fig. S1, E and F). Analysis of the pairwise interactions of the Arabidopsis SNAREs in Arabidopsis epidermal cells using the BiFC system revealed a comparable interaction pattern as for the barley SNAREs (Supplemental Fig. S4, D-F).

Coexpression of Three Complementary SNARE Partners
Next, we coexpressed all three SNARE partners in barley leaf epidermal cells. In this experiment, one SNARE protein was expressed as a translational fusion to CrFP and the other two as a split YFP couple (translationally fused to N YFP and C YFP, respectively). In all tested combinations, the three SNARE partners completely colocalized in large clusters of VLCs (Fig.  2, G-I), suggesting that intermolecular interactions promote and/or stabilize higher order SNARE protein assemblies. The size of the VLC clusters harboring all three SNARE partners was variable and ranged from less than 1 mm to approximately 10 mm. We noted no preferential site at which VLC clusters formed inside the barley epidermal cells. To investigate the nature of the SNARE protein-induced VLCs further, we coexpressed a set of barley proteins known to reside in the PM as mCherry-tagged (Shu et al., 2006) fusion proteins with CrFP-HvVAMP721, N YFP-HvROR2, and HvSNAP34-C YFP in barley epidermal cells. This revealed that all tested barley integral membrane proteins coaccumulated, to varying degrees, with the SNAREs at or in the VLC clusters, suggesting that the majority of PM-targeted proteins may completely or partially accumulate at or in the SNARE-induced VLCs (Supplemental Fig. S3).
The fact that only a few tested combinations (e.g. N YFP-HvROR2 coexpressed with C YFP-HvVAMP721) provided a significant BiFC signal already shows that BiFC formation is very specific and requires the two halves to be in close proximity and a favorable orientation. To further exclude that unspecific BiFC signals can be induced by high local concentration of the fluorescent halves, we coexpressed barley cytochrome b561 (a non-SNARE protein coaccumulating in VLCs; Supplemental Fig. S3, M-P) as a BiFC partner (HvCytb561-C YFP) with N YFP-HvROR2 and CrFP-HvVAMP721. This combination also induced the formation of VLCs, but without any detectable YFP signal (Supplemental Fig. S3, Q-S), which suggests that the N-and C-terminal halves of the split YFP molecule do not even complement at high local concentrations of labeled PM-resident proteins in the VLCs and do not elicit unspecific BiFC signals.
As in case of their barley counterparts, coexpression of CrFP-AtVAMP722, N YFP-AtPEN1, and AtSNAP33-C YFP resulted in coaccumulation of the three SNARE partners in randomly distributed VLC clusters of variable size (Supplemental Fig. S4, G-I). This indicates that the formation of SNARE-induced polypeptide clusters is not a species-specific phenomenon but that the formation of VLC clusters can occur in cells of monocotyledonous as well as dicotyledonous plants.
A membrane-targeted form of CrFP, generated by fusing a CAAX box (CTIL; Thompson and Okuyama, 2000) to the fluorophore, localized next to other cellular compartments and the PM to the VLCs, which suggests that abundant membrane material is present in the VLCs (Supplemental Fig. S4, N-Q). This supports the notion that these compartments are derived from membrane/vesicle fusion events. Notably, even despite high expression levels of the complementary SNARE partner proteins and their clustering in VLCs, the SNARE proteins retained an authentic posttrans- lational dynamic response to powdery mildew attack by focally coaccumulating below the site of attempted fungal penetration (Fig. 3).
A documented weakness of the BiFC assay is that the complemented YFP, once formed, does not disassemble, which stabilizes the entire protein complex (Hu et al., 2002;Held et al., 2008;Kerppola 2008). To exclude the possibility that protein complex stabilization by the BiFC system caused the observed clustering phenomenon, we coexpressed the three SNARE partners tagged with distinct full-size monomeric fluorophores (CrFP-AtVAMP722, mYFP-AtPEN1, and AtSNAP33-mCherry; for mYFP, Zacharias et al., 2002;for mCherry, Shu et al., 2006) in Arabidopsis epidermal cells. Coexpression of these fusion proteins resulted in an accumulation pattern of all three SNAREs in clusters of VLCs that were indistinguishable from the BiFC-associated VLCs (Supplemental Fig. S4, J-M). This demonstrates that the formation of the VLCs and the colocalization of the SNARE proteins are not caused by the use of the BiFC system.

Visualization of in Vivo Ternary SNARE Complex Formation by Combined BiFC-FRET-FLIM Analysis
To further characterize the reconstituted YFP fluorophore, we measured and compared the fluorescence emission spectra and fluorescence lifetime of YFP, YFP N-terminally fused to HvROR2 (YFP-HvROR2), and the reconstituted YFP of the N YFP-HvROR2 and HvSNAP34-C YFP BiFC couple expressed in barley leaf epidermal cells. Consistent with previous publications (Hu et al., 2002), the fluorescence emission spectra of translationally fused and reconstituted YFP were indistinguishable from native YFP (Supplemental Fig.  S5A). In the FLIM experiment, the fluorescence decay curves were fitted with a monoexponential decay model (Kremers et al., 2006). We found an average fluorescence lifetime for YFP of t c m = 3.06 6 0.06 ns (mean 6 SD), which is similar to published values (Kremers et al., 2006). For YFP-HvROR2, t c m was 2.92 6 0.09 ns, and for reconstituted YFP, t c m was 2.74 6 0.08 ns (Table I; Supplemental Fig. S5B). Given that the barrel structure of native YFP and the associated YFP halves might be divergent, these values are surprisingly similar and in the range of the lifetime difference between CrFP and enhanced cyan fluorescent protein, which differ by only three amino acids (Rizzo et al., 2004). Taken together, these data suggest that the reconstituted YFP fluorophore is likely to function as a suitable acceptor molecule.
To assess if all three SNARE partners were in molecular proximity to each other and formed a genuine ternary SNARE complex, we tested if FRET occurred between CrFP donor molecules and different complemented YFP acceptor molecules (for variations in complex formation where FRET, BiFC, or BiFC-FRET can or cannot occur, see scheme in Fig. 4A). The extent of FRET was determined by FLIM (for details, see "Materials and Methods"). For each series of experiments, we determined the average fluorescence lifetime t c m of the donor in the absence of an acceptor (negative control) and compared the result with t c m of the donor coexpressed with various BiFC acceptor combinations.
A monoexponential decay model was sufficient to fit the CrFP fluorescence decay and therefore was used for the analysis of the donor-alone measurements (Supplemental Materials and Methods S1). t c m for CrFP-HvVAMP721 in barley leaf epidermal cells was 2.49 6 0.09 ns (mean 6 SD; Fig. 4, B, C, and J; Table I). However, to adequately fit the fluorescence decay observed upon coexpression of CrFP-HvVAMP721 with N YFP-HvROR2 and HvSNAP34-C YFP, a biexponential model was used (Fig. 4, D-F and J; Table I). The fit yielded t c m = 2.16 6 0.19 ns, which is significantly lower than the control values, indicating effective energy transfer between the CrFP donor molecule and the reconstituted split YFP acceptor molecule.
Analysis of the lifetime decay curves with a monoexponential decay model for both the donor-alone and donor-plus-acceptor studies also revealed a significant obtained from FLIM measurements in barley epidermal cells expressing CrFP-HvVAMP721 alone (blue) CrFP-HvVAMP721 with N YFP-HvROR2 and HvSNAP34-C YFP (red), and CrFP-HvVAMP721 with N YFP-HvROR2 and C YFP-HvSNAP34 (green). Donor-alone fluorescence decay curves were fitted using a monoexponential model, whereas the fluorescence decay in the presence of an acceptor fluorophore had to be fitted with a biexponential model. The second lifetime component was fixed to the average donor lifetime measured in control experiments. The lifetime histograms were generated using a bin size of 100 ps. t c m reduction for CrFP-HvVAMP721 with N YFP-HvROR2 and HvSNAP34-C YFP (2.19 6 0.17 ns) compared with CrFP-HvVAMP722 alone as well (Supplemental Table S1). This control excludes that the observed lifetime reduction is caused by the chosen decay model. The variation in fluorescence lifetime (t m ) throughout an individual cell (Fig. 4F) and the variation in lifetimes t c m averaged over all measured cells (Table I) likely reflect the dynamic nature of the interaction between the three SNARE partners. From the lifetime data, an average FRET efficiency of 12.9% 6 5.7% (mean 6 SD; Eq. 2 in "Materials and Methods") was calculated.
To assess if the reduction of the donor lifetime was only caused by FRET and not by any other photophysical phenomena, such as quenching or photoconversion, APB experiments were performed in a small area of the cell followed by a second FLIM measurement in the same region (Fig. 5, A-G). After APB, YFP fluorescence was almost undetectable (compare the marked areas in Fig. 5A  From the increase in CrFP intensity in this example, a FRET efficiency of 11.2% (Eq. 1 in "Materials and Methods") was calculated, which is similar to the FRET efficiency determined from the lifetime values (see above and Table I). APB measurements on multiple cells yielded an average FRET efficiency of 13.2% 6 7.1% (n = 10) and showed, similar to the lifetime values, considerable variation between individual cells and experiments. Thus, consistent with other studies (Orthaus et al., 2008), APB yielded similar results as FLIM. Hence, the more widely available APB method can be used as an alternative to FLIM to determine FRET between CrFP and a reconstituted YFP.
Following APB, also the CrFP lifetime in the bleached area (t b m ) increased (compare marked areas in Fig. 5, F and G; Supplemental Fig. S5E). Average lifetimes t b m increased from 2.22 6 0.11 ns to 2.43 6 0.04 ns after APB in the bleached area. The average postbleach lifetime differed significantly from prebleach values (t c m of 2.51 6 0.09 ns [P , 0.001]; Table  II) and represents a nearly complete recovery to donoralone values. In conclusion, this experiment confirms that the observed decrease of donor fluorescence lifetime can be attributed to FRET and not to any other photophysical effect that might equally cause a decrease of the donor lifetime.
Positioning of the split C YFP tag to the N terminus instead the C terminus of HvSNAP34 ( C YFP-HvSNAP34) and coexpression with CrFP-HvVAMP721 and N YFP-HvROR2 resulted in lifetimes similar to the donor molecule alone (t c m = 2.44 6 0.13 ns; Fig. 4, G-I; Table I), indicating the absence of FRET. However, cells expressing C YFP-HvSNAP34 and N YFP-HvROR2 showed clearly visible YFP fluorescence, which excludes the possibility that the absence of a suitable  acceptor molecule is the reason for the lack of FRET. The presence of FRET in the former and the absence of efficient FRET in the latter combination of the three SNARE partners are further illustrated by representative decay curves and donor lifetime histograms (Fig. 4, J and K).
To assess SNARE complex formation further, we swapped donor and acceptor molecules in the FRET measurements. When CrFP-HvROR2 was combined with the fluorescent BiFC couple N YFP-HvVAMP721 and HvSNAP34-C YFP, only a small reduction in t c m (2.42 6 0.05 ns) was observed compared with the donor alone (2.50 6 0.09 ns; Table I Table I; Supplemental Fig. S6, F-J). The absence of a lifetime reduction upon changing the position of the separate YFP parts or an exchange of the CrFPlabeled donor molecules was likely the consequence of unfavorable distances or orientations between the fluorescent labels (Fig. 4A, scenario 5). However, this finding supports the notion that the t c m reduction observed in cells expressing CrFP-HvVAMP721, N YFP-HvROR2, and HvSNAP34-YFP C is due to molecular interactions and not molecular crowding, CrFP bleaching, or other phenomena.

Visualization of the Arabidopsis AtPEN1, AtSNAP33, and AtVAMP722 Ternary SNARE Complex in Vivo by BiFC-FRET-FLIM
To examine the applicability of the BiFC-FRET-FLIM approach in another plant species, we also studied complex formation between CrFP-AtVAMP722, N YFP-AtPEN1, and AtSNAP33-C YFP in Arabidopsis. The three proteins were coexpressed in Arabidopsis epidermal cells, and FLIM measurements were performed. The donor CrFP-AtVAMP722 alone had an average fluorescence lifetime t c m of 2.55 6 0.10 ns (Table I; Supplemental Fig. S7B). When CrFP-HvVAMP721 was coexpressed with the acceptor couple N YFP-AtPEN1 and AtSNAP33-C YFP, which as in the case of the respective barley orthologs led to reconstitution of YFP, t c m decreased to 2.24 6 0.18 ns (Table I; Supplemental Fig. S7, C-E). These results show that BiFC-FRET-FLIM can also be applied to study ternary SNARE complex formation in Arabidopsis leaf epidermal cells and suggest that the method is applicable to a broad range of plant species.

BiFC Combined with FRET Reveals the Formation of Ternary SNARE Complexes in Living Plant Cells
In this work, we applied BiFC and BiFC combined with FRET to study the pairwise and ternary interactions, respectively, between a set of complementary barley and their orthologous Arabidopsis SNARE proteins that were previously shown to be essential in disease resistance and plant development (Heese et al., 2001;Collins et al., 2003;Assaad et al., 2004;Zhang et al., 2007;Kwon et al., 2008;Pajonk et al., 2008). Our results corroborate the biochemical detection of these SNARE complexes ( Fig. 1; Kwon et al., 2008;Pajonk et al., 2008) and indicate that the biochemically isolated SNARE complexes were not artificially formed during the protein extraction procedure.
To investigate the formation of ternary SNARE complexes, FLIM measurements were performed in epidermal leaf cells coexpressing two SNAREs as BiFC partners (reconstituting a functional YFP acceptor molecule) and the third SNARE fused to CrFP as FRET donor. In the cases where CrFP-HvVAMP721 was used as donor molecule with a YFP reconstituted from N YFP-HvROR2 and HvSNAP34-C YFP as acceptor, a significant decrease of the fluorescence lifetime (t c m ), indicative of effective FRET, was observed ( Fig.  4F; Table I). Similar results were obtained in the Arabidopsis system (Table I; Supplemental Fig. S7E). However, a significant decrease of t c m only occurred when a C-terminal fusion of HvSNAP34 to C YFP was Table II. Fluorescence lifetime and energy transfer efficiency determined after photobleaching of the acceptor fluorophore To verify that the observed decrease of fluorescence lifetime was only due to FRET, the complemented YFP was bleached in a selected area and the fluorescence lifetimes before and after bleaching were determined. At the same time, the APB experiments served as an independent method to determine the FRET efficiency between the CrFP donor and complemented YFP acceptor molecules. used, while we observed no FRET when an N-terminal fusion of HvSNAP34 to C YFP was tested ( Fig. 4I; Table  I). This result may indicate that the two SNARE domains of HvSNAP34 preferentially adopt an antiparallel orientation (Weninger et al., 2003). Additionally, when either CrFP-HvROR2 or HvSNAP34-CrFP was used as a donor molecule, no reduction in t c m occurred (Table I; Supplemental Fig. S6, E and J). When conventional (full-size) fluorophores are employed, FRET can be typically observed in reciprocal experiments with the donor and acceptor fluorophores swapped between the interacting partners (Russinova et al., 2004). However, next to the distance, FRET depends on the dipole orientation of the fluorophores. The use of the BiFC system limits the rotational freedom of the YFP molecule, because the two YFP halves are attached to two interacting partners. This decreases the chance that the CrFP and YFP molecules are properly aligned for FRET to occur (Dale et al., 1979;Chen et al., 2003). By switching the position of the YFP halves fused to the SNAREs, the position of the "fixed" YFP moiety changes, which may provide a plausible explanation for the fact that we did not detect FRET in certain donor/acceptor combinations (Fig. 4A, scenarios 4 and 5). These requirements for fluorophore positioning and orientation as well as the limited rotational freedom of reconstituted YFP suggest that FRET efficiencies determined in a BiFC-FRET three-component system will possibly be lower than for a binary FRET assay.
The finding that the FRET efficiency determined for the ternary SNARE complex is comparable to the efficiency for respective binary SNARE complexes (Table I; Kwon et al., 2008) indicates that the capacity of complemented YFP to serve as a FRET acceptor can equal that of native YFP under favorable conditions. APB confirmed the results of the fluorescence lifetimebased FRET assays. By performing a second FLIM measurement after bleaching, we showed that the lifetime (t b m ) recovered significantly in the bleached area, which demonstrates that the lifetime (t c m ) reduction is due to FRET and not caused by any other quenching process or photoconversion of the fluorophore (Biskup et al., 2004;Tramier et al., 2006;Hoffmann et al., 2008). The fact that the t c m did not completely recover to donor-alone values is possibly due to partly bleached and/or photoconverted CrFP donor molecules, which can occur during the imaging and APB procedure .
Extending the basic bimolecular FRET setup by tagging molecules with three spectrally overlapping fluorophores has been employed by others to study ternary molecule interactions. This approach, for example, was used to study the interactions and the composition of complexes composed of small molecules in vitro (Lee et al., 2007), to follow molecular interactions over distances greater than 10 nm (Haustein et al., 2003), and to visualize interactions between three genetically encoded fluorophore-tagged fusion proteins in vivo (Galperin et al., 2004). The fluorescence intensity-based FRET approaches used in these studies rely on alternating excitation of the fluorophores and subsequent measurements of fluorescence intensities in separate channels to distinguish direct FRET from two-step FRET between the partners under study. Extensive calculations, taking the various FRET contributions and corrections for spectral bleed-through into account, then provide information about intermolecular distances and complex composition. Another method combines bioluminescence resonance energy transfer with BiFC to follow the formation of protein complexes. Due to the low luminescence and fluorescence signals, this method cannot be used to study phenomena at (sub)cellular resolution (Gandia et al., 2008). Recently, BiFC has been used in combination with FRET to visualize the ternary Fos-Jun-nuclear factor of activated T cells (NFAT) complex in mammalian COS-1 cells. This study, which paralleled our own experiments, demonstrated the formation of the heterooligomeric NFAT complex and the establishment of a ternary protein complex between Fos-Jun heterodimers and the NF-kB subunit p65 (Shyu et al., 2008a(Shyu et al., , 2008b. Here, we further corroborated the validity of this approach by employing FLIM, which is a more precise method to determine FRET and, unlike intensity-based approaches such as sensitized emission and APB, independent of fluorophore concentrations .

Practical Implications for Conducting BiFC-FRET Measurements in Plant Cells
BiFC-FRET determination by APB or FLIM has caveats that should be considered when setting up a three-protein-interaction study. The drawbacks of using the BiFC system to monitor protein-protein interactions were discussed above and elsewhere (Hu et al., 2002;Held et al., 2008;Kerppola, 2008) and include, among others, the irreversibility of complex formation, which may stabilize specific as well as unspecific protein-protein interactions. Another implication of the irreversibility of BiFC complex formation is that, with respect to two of the partners, ternary protein complexes can only be studied statically by BiFC-FRET analysis. Owing to the high energy that is required to dissociate a BiFC complex (Kerppola, 2009), dynamic assemblies or disassemblies of the two BiFC partners cannot be resolved. In addition, high local concentrations of proteins in organelles or in the PM might promote the formation of BiFC complexes and therefore possibly visualize false-positive protein-protein interactions. A major drawback of FRET studies in general is that high numbers of photons are needed to reliably determine energy transfer. Therefore, relatively high concentrations of fluorophores, often only reached in an overexpression system, are required. These nonnatural concentrations might lead to ectopic protein localization or to altered cellular responses that may promote interactions between proteins and FRET due to molecular crowding.
For these reasons, FRET-based methods should include the proper (e.g. noninteracting) controls and should ideally be backed up by independent methods that demonstrate a protein-protein interaction (e.g. coimmunoprecipitation, far western blot, etc.). Ideally, BiFC-FRET measurements would be conducted using native expression levels (all three genes driven under control of their endogenous promoters) in a respective triple mutant background. For practical reasons, this "gold standard" will be impossible or impractical to obtain in most instances. Instead, nonnative transient gene expression, either mediated by Agrobacterium tumefaciens infiltration or protoplast transformation or following particle bombardment (as in this study), will remain common practice for fluorophore-based interaction studies (Bhat et al., 2006).
In addition, all FRET-based methods are prone to photophysical artifacts, like bleaching or photoconversion of the donor or acceptor molecule by prolonged exposure to light, which might not only change the intensity but may also reduce the lifetime of fluorophores with a multiexponential fluorescence decay such as CrFP and cyan fluorescent protein (Kremers et al., 2006;Tramier et al., 2006;Hoffmann et al., 2008;Villoing et al., 2008). This requires again proper controls or an independent method to confirm the interaction. One possibility is to combine FLIM with APB, as demonstrated in this study (Fig. 5;Supplemental Fig. S5E). A major advantage of demonstrating FRET between a CrFP and reconstituted YFP by FLIM compared with intensity-based approaches (Galperin et al., 2004;Shyu et al., 2008aShyu et al., , 2008b is that the method is independent of fluorophore concentrations and provides a more precise quantification of the interaction. On the other hand, it requires sophisticated equipment, which is not always available. As successfully shown here and by others, APB or sensitized emission measurements can be used to demonstrate BiFC-FRET (Shyu et al., 2008a(Shyu et al., , 2008b, which makes the method accessible to a wider range of the plant community.

Overexpression of Complementary SNARE Proteins Results in the Formation VLC Clusters
Complementary to our FRET studies, the presence of a ternary SNARE complex containing HvROR2 was demonstrated by biochemical means (Fig. 1). This shows by an independent method that HvROR2, like AtPEN1 in Arabidopsis , forms ternary SNARE complexes in planta. With the tools available, it was impossible to demonstrate the presence of HvSNAP34 or HvVAMP721 in this complex. The visualized complexes may thus represent a mixture of the ternary SNARE complex composed of HvSNAP34, HvVAMP721, and HvROR2 and other HvROR2-containing SNARE complexes.
Coexpression of two or three SNARE partners fused to complementary halves of a split YFP molecule revealed the formation of VLC clusters that contain the respective interacting SNARE partners (Fig. 2, D-J). The VLCs are not present in cells expressing only one of the SNARE partners and therefore are the consequence of the co-overexpression of two or more distinct SNAREs. They colocalize with a CrFP-CAAX membrane marker, and a range of PM-resident proteins are targeted to the same location. It thus seems that PM proteins together with membrane material in transit through the secretory pathway get targeted to the VLCs by the action of the co-overexpressed SNAREs.
Notably, VLC formation was not influenced by the BiFC system (Supplemental Fig. S4, J-M). Thus, the formation of a complemented YFP moiety does not significantly inhibit vesicle fusion events, suggesting that the BiFC complexes retain the ability for the formation of functional ternary SNARE complexes. When coexpressed and challenged with Bgh, the fluorescent SNARE fusion proteins focally accumulated below the fungal appressorium and the primary germ tube (Fig. 3). Occasionally, the accumulation sites were not shaped as a sphere but as scattered spots surrounding the fungal appressorium (Fig. 3, E-H). This resembles the pattern observed for the accumulation of hydrogen peroxide upon fungal challenge, where patches of hydrogen peroxide-containing papillae surround the central wall apposition . In both cases, the patterns of accumulation are likely the result of transport processes toward fungal attack sites.
At present, it is unclear what the origin of the VLCs is. One possibility is that the overexpression of multiple SNAREs results in runaway vesicle fusion due to disequilibrium between vesicle fusion and dissociation by exceeding the capacity of the SNARE complexresolving machinery. This might artificially increase the fraction of interacting SNARE partners in our experimental system. It is conceivable that proteins that are normally cotransported with the SNAREs (e.g. on or inside vesicles) would coaccumulate at the same locations during this process. Alternatively, the overexpression of complementary exocytotic SNAREs may trigger a chain of cellular events that otherwise only occurs upon a biotic stress stimulus. With respect to their molecular composition, the VLC clusters could thus represent either an intracellular congregation of PM-resident proteins and vesicular compartments into an extremely large multiprotein complex (PM microdomain; Bhat et al., 2005) or the exaggerated formation of multivesicular bodies (Meyer et al., 2009).

CONCLUSION
Taken together, our findings demonstrate the technical feasibility of combined BiFC-FRET-FLIM and BiFC-FRET-APB measurements to monitor the existence of ternary protein complexes in living plant cells. We envisage future variations of this basic setup by the employment of multicolor fluorescence complementation analysis (Hu and Kerppola 2003;Jach et al., 2006;Lee et al., 2008;Waadt et al., 2008) combined with FRET to visualize higher order complex formation. By choosing proper combinations of interacting proteins and respective split fluorophores and applying a triple FRET method as described by Galperin et al. (2004) or spectral/multichannel FLIM complexes, up to five interacting proteins can possibly be visualized.

DNA Constructs
The expression vectors used in this study were constructed using Gateway technology (Invitrogen; www.invitrogen.com), unless stated otherwise. The pUCSPYNE, pUCSPYCE, pESPYNE, and pESPYCE vectors harboring N-and C-terminal halves of enhanced YFP (Walter et al., 2004;Schü tze et al., 2009) were adapted to be Gateway compatible.

Microscopy
Fluorescence microscopy was carried out with a Zeiss LSM510 META confocal microscope (www.zeiss.com). A 403 Plan-Apochromat (Zeiss) water-immersive lens with a numerical aperture of 1.2 was used. CrFP or YFP was excited with the 458-or 514-nm laser line of an argon-ion laser, respectively. Fluorescence and excitation light were separated by a dual dichroic mirror, which reflected both the 458-and 514-nm laser lines. A second dichroic NFT515 filter was used to split the emission light into the CrFP and YFP detection channels. The CrFP fluorescence was selected by a 480-to 520-nm bandpass filter, and YFP fluorescence was recorded in META detector channels corresponding to the wavelength range from 526 to 569 nm. mCherry and FM 4-64 were excited by the 543-nm laser line of the helium-neon laser. Excitation and emission light were separated by a 543-nm dichroic filter. mCherry fluorescence was selected by a 565-to 615-nm bandpass filter, and FM 4-64 fluorescence was filtered by a 650-to 710-nm bandpass filter. To better separate the fluorescence of the coexpressed fluorophores, the measurements were performed in independent tracks exciting only one fluorophore at a time.

BiFC-FRET-APB
The same beam paths as mentioned above were used. The bleaching was performed as a time series with three image scans before bleaching and three or more images after bleaching. YFP was bleached by iteratively scanning a selected area of the cell 80 times with the 514-nm laser line of the argon-ion laser at a laser power of 178 mW in the focal plane and a pixel dwell time of 2.6 ms. The images were analyzed with Image J 1.42 (National Institutes of Health). Background signal was subtracted, and the FRET efficiency was calculated according to where I DA is the intensity of the donor in the presence of the acceptor (i.e. before bleaching of the acceptor) and I D is the intensity of the donor in the absence of the acceptor (i.e. after bleaching of the acceptor; for details, see Supplemental Materials and Methods S1). An average FRET efficiency was calculated from the three images taken immediately after the bleach.

Fluorescence Lifetime Microscopy
Fluorescence lifetime imaging microscopy was essentially performed as described previously (Bhat et al., 2005;Shen et al., 2007). In brief, a modelocked titanium-sapphire laser (Chameleon; Coherent) was tuned to a wavelength of 840 nm for two-photon excitation of CrFP and tuned to 910 nm for YFP excitation. A dichroic 650KP beam splitter was used to separate excitation and emission light. Emission light was directed to the NDD port of the confocal laser scanning microscope. A BG39 filter (Schott) was used to block reflected excitation light, and a 465-to 495-nm or a 540-to 580-nm bandpass filter was used to select CrFP or YFP fluorescence, respectively. Single-photon events were detected with a multichannel plate photomultiplier (R3809U-52; Hamamatsu Photonics), which was connected to a SPC730 module (Becker and Hickl; www.becker-hickl.com) for time-correlated single-photon counting. For image acquisition, the area of interest was scanned continuously. The time-correlated single-photon-counting module was synchronized to the scan via the FrameSync, LineSync, and PixelSync signals of the confocal laser scanning microscope. The temporal resolution of this setup was better than 50 ps (full width half maximum).
An acquisition time of 120 s was sufficient to accumulate histograms of the fluorescence decay with a resolution of 128 3 128 pixels and 256 time channels. The histograms were analyzed with SPC-Image 2.9.1 software (Becker and Hickl) and our own software routines written in MATLab (MathWorks; for more details, see Supplemental Materials and Methods S1). Both software versions used the "iterative reconvolution method" to fit a monoexponential or biexponential decay function to the data. The parameters were approximated by a modified Levenberg-Marquardt algorithm, and the goodness of a fit was judged by the value of the reduced x 2 v . To limit the number of parameters in a biexponential decay function to be fit, the second lifetime component was fixed to the average donor lifetime found in the donor-only measurements performed during the same session. A fit was regarded to be acceptable when x 2 v , 1.2. To visualize the lifetime distribution over the cell, average lifetimes were color coded.
In the case of a biexponential fit, amplitude-weighted mean lifetimes (t m ) were calculated according to t m = A f t f + A s t s , where A f and A s are the relative amplitudes of the fast and slow lifetime components, respectively. Mean lifetimes (t c m ) for an individual cell were calculated by averaging over the amplitude-weighted mean lifetimes (t m,i ) of all pixels (i) series of measurements (j) are presented as means 6 SD based on more than 10 cells from three or more independent experiments. In contrast to a monoexponential decay model, the use of a biexponential decay model consistently yielded x 2 v , 1.2 for the combinations showing FRET and therefore is regarded as the proper model to describe the lifetime decay data.
By comparing the lifetime of the donor in the presence of the acceptor t c DA with the lifetime measured in the absence of the acceptor t c D , the FRET efficiency can be calculated: As fluorescence lifetimes are independent of the concentration of the fluorophores, FRET and control measurements can be carried out in different cells (for more details, see Supplemental Materials and Methods S1).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Additional details of subcellular SNARE localization.
Supplemental Figure S2. Results of screening experiments for fluorescent complexes of SNARE proteins fused to complementing halves of a split YFP.
Supplemental Figure S3. Coexpression of the three complementary barley SNARE partners with HvMLO, HvLTP, HvPIP2, or HvCytb561 leads to association of these proteins to VLCs.
Supplemental Figure S4. Coexpression of the complementary Arabidopsis SNAREs AtPEN1, AtSNAP33, and AtVAMP722 also results in association of the SNARE proteins to VLCs.
Supplemental Figure S5. Exemplary fluorescence emission spectra and fluorescence decay curves.
Supplemental Figure S6. Exchanging the donor molecule in the ternary SNARE complex abolishes FRET.
Supplemental Figure S7. In planta visualization of ternary SNARE complex formation by combined BiFC-FRET-FLIM in Arabidopsis leaf epidermal cells.
Supplemental Table S1. Fluorescence lifetimes and energy transfer efficiencies using a monoexponential fit.
Supplemental Materials and Methods S1.