Manipulating Endoplasmic Reticulum-Plasma Membrane Tethering in Plants Through Fluorescent Protein Complementation

The bimolecular fluorescence complementation (BiFC) assay has been widely used to examine interactions between integral and peripheral proteins within putative plasma membrane (PM) microdomains. In the course of using BiFC assays to examine the co-localization of plasma membrane (PM) targeted receptor-like kinases (RLKs), such as FLS2, with PM micro-domain proteins such as remorins, we unexpectedly observed heterogeneous distribution patterns of fluorescence on the PM of Nicotiana benthamiana leaf cortical cells. These patterns appeared to co-localize with the endoplasmic reticulum (ER) and with ER-PM contact sites, and closely resembled patterns caused by over-expression of the ER-PM tether protein Synaptotagmin1 (SYT1). Using domain swap experiments with SYT1, we inferred that non-specific dimerization between FLS2-VenusN and VenusC-StRem1.3 could create artificial ER-PM tether proteins analogous to SYT1. The same patterns of ER-PM tethering were produced when a representative set of integral membrane proteins were partnered in BiFC complexes with PM-targeted peripheral membrane proteins, including PtdIns(4)P-binding proteins. We inferred that spontaneous formation of mature fluorescent proteins caused the BiFC complexes to trap the integral membrane proteins in the ER during delivery to the PM, producing a PM-ER tether. This phenomenon could be a useful tool to deliberately manipulate ER-PM tethering or to test protein membrane localization. However, this study also highlights the risk of using the BiFC assay to study membrane protein interactions in plants, due to the possibility of alterations in cellular structures and membrane organization, or misinterpretation of protein-protein interactions. A number of published studies using this approach may therefore need to be revisited.

The bimolecular fluorescence complementation (BiFC) assay has been widely used to examine interactions between integral and peripheral proteins within putative plasma membrane (PM) microdomains. In the course of using BiFC assays to examine the co-localization of plasma membrane (PM) targeted receptor-like kinases (RLKs), such as FLS2, with PM micro-domain proteins such as remorins, we unexpectedly observed heterogeneous distribution patterns of fluorescence on the PM of Nicotiana benthamiana leaf cortical cells. These patterns appeared to co-localize with the endoplasmic reticulum (ER) and with ER-PM contact sites, and closely resembled patterns caused by over-expression of the ER-PM tether protein Synaptotagmin1 (SYT1). Using domain swap experiments with SYT1, we inferred that non-specific dimerization between FLS2-VenusN and VenusC-StRem1.3 could create artificial ER-PM tether proteins analogous to SYT1. The same patterns of ER-PM tethering were produced when a representative set of integral membrane proteins were partnered in BiFC complexes with PM-targeted peripheral membrane proteins, including PtdIns(4)P-binding proteins. We inferred that spontaneous formation of mature fluorescent proteins caused the BiFC complexes to trap the integral membrane proteins in the ER during delivery to the PM, producing a PM-ER tether. This phenomenon could be a useful tool to deliberately manipulate ER-PM tethering or to test protein membrane localization. However, this study also highlights the risk of using the BiFC assay to study membrane protein interactions in plants, due to the possibility of alterations in cellular structures and membrane organization, or misinterpretation of protein-protein interactions. A number of published studies using this approach may therefore need to be revisited.

SIGNIFICANCE STATEMENT
The bimolecular fluorescence complementation (BiFC) assay has been widely used to examine interactions between integral and peripheral proteins within putative plasma membrane (PM) microdomains. Using domain swap experiments involving the endoplasmic reticulum-PM tether protein SYT1, we have obtained evidence that BiFC complexes involving one integral membrane protein and one peripheral membrane protein can act as synthetic EM-PM tethers, producing images that resemble microdomain co-localization, but are actually artifacts; a number of published studies should thus be revisited.

INTRODUCTION
close to the TMD, and a C-terminal cytoplasmic domain containing two conserved calcium-binding domains C2A and C2B that are responsible for binding a variety of negatively charged phospholipids on the PM (Schapire et al., 2008;Yamazaki et al., 2010;Levy et al., 2015;Pérez-Sancho et al., 2015). The TMD mediates ER-anchoring and the C2AC2B domain mediates PM-binding, together conferring the ER-PM tethering function of SYT1. Plant EPCSs have also been shown to contain VAP27-1 and VAP27-3 which are homologous to mammalian VAPs (Wang et al., 2014(Wang et al., , 2016. The C-terminal domains of the VAP27s are examples of integral membrane tail-anchoring domains, and anchor the VAP27s to the ER membrane. PM-binding of the VAP27s is via interaction of their N-terminal conserved major sperm domain (MSD) with the NETWORKED protein, NET3C, as well as cytoskeletal elements. NET3C is plant-specific and located in EPCSs. In addition, VAP27s were also found to form protein complexes with plant oxysterol-binding-protein-related proteins (ORPs), which contain pleckstrin homology (PH) domains for the binding of phosphoinositides (PIPs) in the PM (Saravanan et al., 2009). Furthermore, VAP27 was recently shown to directly interact with clathrin and with PIPs at endocytic subdomains in the PM, resulting in the establishment of tethering (Stefano et al., 2018).
The bimolecular fluorescence complementation (BiFC) assay (Kerppola, 2008) is a commonly used experimental approach to study protein-protein interactions (Kerppola, 2008). The BiFC assay is based on two non-fluorescing fragments split from a fluorescent protein, each of which is translationally fused with a different protein of interest; interaction between the proteins of interest will bring the two non-fluorescing fragments into proximity with each other resulting in the re-assembly of a functional fluorescent protein (Kerppola, 2006). Thus, the BiFC assay not only enables identification of a potential proteinprotein interaction, but also allows direct visualization of the protein complex in vivo. Due to these useful characteristics, the BiFC assay has also been successfully applied as a highthroughput approach in several large-scale studies to map potential protein-protein interactions (Remy and Michnick, 2004;Boruc et al., 2010;Snider et al., 2013). Venus, a variant of enhanced yellow fluorescence protein (EYFP) with a higher efficiency of maturation and better adaptability in acid and high temperature environments, has become a widely used fluorescent protein for BiFC assays (Saka et al., 2007;Kodama and Hu, 2010;Miller et al., 2015). Moreover, a residue at position 155 has proven useful as a split site for Venus in BiFC assays (Wong and O'Bryan, 2011;Kodama and Hu, 2012). However, a challenge for this strategy is the spontaneous reassembly of the two fragments in the absence of associating protein partners that can result in false positive BiFC signals (Shyu et al., 2006;Zamyatnin et al., 2006;Saka et al., 2007).
Growing evidence has revealed that different phospholipid species and membrane proteins in the PM may be organized into coalescences with diameters ranging from 2.0 to 300 nm, referred to as "lipid or membrane rafts" or more recently as "microdomains" (Kusumi et al., 2011;Lillemeier and Klammt, 2012;Varshney et al., 2016). These microdomains are the result of lipid-lipid, protein-lipid, and protein-protein interactions in the plasma membrane, potentially providing functional platforms to orchestrate a multitude of signaling pathways (Kusumi et al., 2012). To date, two major protein families, called flotillins and remorins have been associated with plasma membrane microdomains (Raffaele et al., 2009). Flotillins are widely present in all kingdoms of life, and their membrane targeting is mediated by either myristoylation, palmitoylation, or both (Jarsch et al., 2014). In contrast, remorins are plant-specific proteins, which have been well-characterized and contain a highly conserved Cterminal coiled-coil domain for plasma membrane anchoring (Perraki et al., 2012).
In plants, a spectrum of PM-bound receptor-like kinases (RLKs) are employed to coordinate signaling pathways in growth, development, and innate immunity (He et al., 2018). Several RLKs have been found to be functionally associated with remorins or flotillins (Jarsch et al., 2014). For example, the remorin protein MtSYMREM1 from the legume Medicago truncatula, was reported to function as a scaffolding protein mediating spatial distribution of several RLKs during symbiotic plant-microbe interactions (Lefebvre et al., 2010), including MtNFP (Arrighi et al., 2006), MtLYK3 (Smit et al., 2007), and MtDMI2 (Limpens et al., 2005). Likewise, the closest Lotus japonicus homolog of MtSYMREM1, LjSYMREM1 (Tóth et al., 2012), was shown to interact with the L. japonicus RLKs, LjNFR5 , LjNFR1 , and LjSYMRK (Stracke et al., 2002). The Arabidopsis flotillin protein, AtFlotillin1 (Borner et al., 2005), was shown to be critically involved in the activation of the RLK growth regulator, AtBRI1 (Russinova et al., 2004); the two proteins showed increased co-localization in response to the brassinosteroid ligand (Wang et al., 2015). More recently, Bücherl et al. (2017) observed that in Arabidopsis, AtBRI1 and the RLK immune receptor AtFLS2 (Gómez-Gómez and Boller, 2000) are heterogeneously but differently distributed in the membrane and that each receptor was associated with distinct remorin proteins. Despite these advances, the underlying mechanisms of compartmentalization of cell surface receptors into plasma membrane microdomains is still incompletely understood.
In this study, we set out to investigate pairwise associations between a set of representative membrane receptors, remorins and lipid-binding proteins using the BiFC assay. When RLKs such as FLS2 were ectopically co-expressed with Solanum tuberosum remorin StRem1.3 in N. benthamiana cortical cells, the BiFC fluorescent signal was heterogeneously distributed in distinct patch-like domains or nearly continuous networks across the PM. Co-localization assays suggested that these patterns may be associated with ER-PM contact sites, and thus that the BiFC complexes might unexpectedly be acting as artificial ER-PM tethering proteins. Here, using domain swap experiments involving the tether protein SYT1, we have obtained evidence that any BiFC complex that involves one integral membrane protein and one peripheral membrane protein has the potential to act as an artificial ER-PM tethering protein. This artifact has been overlooked in previous studies of membrane organization using BiFC assays.

Heterogeneous Patch-Like Distribution of FLS2-StRem1.3 BiFC Complexes
To provide a positive control for our studies of protein-lipid organization in the PM using BiFC, we first chose FLS2 and StRem1.3, which have been reported to co-localize with each other at the haustorial interface when FLS2 is activated by flg22 (Bozkurt et al., 2015). To do this, we fused the N-terminal fragment of Venus (VenusN) to the C terminus of FLS2, and the C-terminal fragment of Venus (VenusC) to the N terminus of StRem1.3. We also made complementary FLS2-VenusC and VenusN-StRem1.3 constructs. The pairs of BiFC constructs were ectopically co-expressed under the control the Cauliflower Mosaic Virus (CaMV) 35S promoter in N. benthamiana leaf cortical cells using Agrobacterium tumefaciens-mediated transient transformations. By confocal microscopy live-cell imaging, we observed strong BiFC fluorescence signals with both BiFC configuration pairs: FLS2-VenusN + VenusC-StRem1.3 and FLS2-VenusC + VenusN-StRem1.3 (Figure 1). In both configurations, the BiFC fluorescence signal was heterogeneously distributed in distinct patterns ranging from discrete patches through to continuous networks spanning the cortical surface. As a control, we co-expressed the FLS2 and StRem1.3 fusions with complementary Venus fragments that were not fused to another protein. In each case, we observed appreciable BiFC fluorescent signals (Supplementary Figure S1), indicating that non-specific interactions between the two fragments of Venus could occur in the absence of FLS2-StRem1.3 associations (Kodama and Hu, 2010;Gookin and Assmann, 2014). However, the BiFC fluorescence signals produced by each of these control pairs were homogenously distributed on the plasma membrane, especially in the case of FLS2. In each case, the subcellular localization closely matched that of fusions of FLS2 or StRem1.3 with full-length YFP (Supplementary Figure S1), indicating that the subcellular localization of each control BiFC complex was determined by the respective FLS2 or StRem1.3 component. It also indicated that the heterogeneous distribution patterns observed with the FLS2-StRem1.3 BiFC complexes were produced only when both FLS2 and StRem1.3 were present. Similar results were obtained when using YFP as the BiFC fluorophore (Supplementary Figure S2), or when the constructs were expressed in Arabidopsis mesophyll protoplasts (Supplementary Figure S2). When StRem1.3 was replaced by a mutant, StRem1.3 * that could not bind the PM, the BiFC complexes displayed the localization expected for FLS2 alone (Supplementary Figures S1B,C). StRem1.3 * contains mutations in the membrane-insertion loop of StRem1.3 that abolish the hydrophobicity of the loop (Perraki et al., 2012).

FLS2-StRem1.3 BiFC Complexes Appear to Localize to ER-PM Contact Sites
Since the net-like distribution of the FLS2-StRem1.3 BiFC complexes in many cells resembled the distribution of the cortical ER, we performed co-localization assays using the ER marker SP-tagRFP-HDEL (Matsushima et al., 2002). The cellular distribution of FLS2-StRem1.3 BiFC fluorescence closely followed the distribution patterns of SP-tagRFP-HDEL, namely net-like and sheet-like patterns of fluorescence (Figure 2A). We also commonly observed dynamic movements of the ER network (Stefano et al., 2014) when SP-tagRFP-HDEL was expressed alone ( Figure 2B, Supplementary Movie S1). We documented the dynamic movements of the labeled organelles by using kymographs, in which the fluorescent signal along a transect is imaged over time. As shown in Figure 2B, the dynamic movements of the ER network labeled by SP-tagRFP-HDEL produced a chaotic kymograph. In contrast, we observed that the puncta of the FLS-StRem1.3 BiFC complexes were relatively static, producing straight lines on the kymograph ( Figure 2B, Supplementary Movie S2). Moreover, in cells co-expressing both FLS2-StRem1.3 BiFC complexes and SP-tagRFP-HDEL, the puncta of the FLS2-StRem1.3 BiFC complexes co-localized with SP-tagRFP-HDEL at junctions in the ER network where the SP-tagRFP-HDEL signal showed increased stability ( Figure 2C). However, small portions of the ER networks that were not colocalized with FLS2-StRem1.3 complexes still showed dynamic movements ( Figure 2C). Since ER-PM contact sites are sites of reduced mobility of the ER (Henne et al., 2015), we hypothesized that the FLS2-StRem1.3 puncta may correspond to ER-PM contact sites.
To test whether the FLS2-StRem1.3 puncta may correspond to ER-PM contact sites, we co-expressed tagRFP-tagged SYT1 protein from Arabidopsis, which has been wellcharacterized as a tethering protein for ER-PM contact sites (Pérez-Sancho et al., 2015). As shown in Figure 2D, FLS2-StRem1.3 BiFC signals were clearly co-localized with SYT1. Moreover, a characteristic property of ER-PM junctions in mammalian cells is that they restrict the distribution of other membrane proteins in mammalian cells (Carrasco and Meyer, 2011). Though the exclusion of membrane proteins by ER-PM junctions has not been reported on plants, when we examined the distribution of membrane-associated protein AtFlotillin1 (Supplementary Figure S3E), reduced distribution in regions of the membrane displaying either SYT1-YFP or especially the FLS2-StRem1.3 complexes was apparent, as revealed by both maximum intensity projection and orthogonal projection (Supplementary Figures S3B,D) whereas general membrane labeling by FM4-64 was not restricted (Supplementary Figures S3A,C).

StRem1.3 and Other Peripheral Membrane Proteins Can Replace the C-Terminus of SYT1 in ER-PM Tethering
The ER-PM tethering protein SYT1 contains an N-terminal ER transmembrane domain (SYT1n) and a C-terminal peripheral PM-binding C2AC2B domain (Prinz, 2014; Figure 3A). As shown in Figure 3B, formation of a Venus BiFC complex was sufficient to reconstitute the membrane tethering function of the separated SYT1 N-and C-terminal domains. The complex showed the same distribution and stability as intact SYT1 ( Figure 3C, Supplementary Figure S4). Removal of the C-terminal C2AC2B domain of SYT1 resulted in a dynamic net-like distribution, whether the SYT1 N-terminus was labeled with full length YFP ( Figure 3D, Supplementary Figure S4) or a Venus BiFC complex ( Figure 3E). This dynamic distribution co-localized with the ER marker SP-tagRFP-HDEL (Supplementary Figure S4). However, adding StRem1.3 to the C-terminus of SYT1n via a BiFC complex was sufficient to restore the stable ER-PM site distribution ( Figure 3F) and could also partially stabilize the distribution of co-expressed SP-tagRFP-HDEL (Supplementary Figure S4). When StRem1.3 was replaced by the mutant StRem1.3 * , the ability to reconstitute ER-PM tethering with SYT1n was abolished ( Figure 3G). In addition, to reduce the possibility that tethering is an artifact of over-expression in the N. To test if other peripheral membrane proteins could also participate in ER-PM tethering, we replaced the C2AC2B domain of SYT1 with the well-studied receptor-like cytoplasmic kinases (RLCKs), BIK1 (Lu et al., 2010), PBS1 (Qi et al., 2014), or CPK21 (Asai et al., 2013); these three proteins are targeted to the PM by either by N-terminal myristoylation, palmitoylation or both (Supplementary Figure S6). Coexpression of BIK1-VenusN + SYT1n-VenusC, PBS1-VenusN + SYT1n-VenusC, and CPK21-VenusN + SYT1n-VenusC all produced stable puncta-like distributions resembling ER-PM tethering (Supplementary Figure S6), which was further confirmed by co-localization analysis using SYT1 fused with tagRFP (Supplementary Figure S6).

Integral Membrane Proteins Can Contribute ER Anchoring to Produce ER-PM Tethering
Integral membrane proteins (IMPs) such as FLS2 are synthesized on the ER, with the N-terminal domain in the lumen of the ER and the C-terminal domain in the cytoplasm (Walter and Johnson, 1994;Goder and Spiess, 2001). We hypothesized that the reason that FLS2 could participate in tethering-like complexes was because the formation of the FLS2-StRem1.3 complex trapped FLS2 in the ER, with its C-terminal-attached VenusN or VenusC fragment in the cytoplasm, bound to the StRem1.3 component ( Figure 4A). As demonstrated above (Figure 1) To address whether other IMPs could also form this ER-PM tethering structure with StRem1.3 through BiFC self-assembly, we selected several well-studied plasma membrane RLKs which share similar structural characteristics with FLS2 and also have similar localization patterns (Supplementary Figure S7). RLKs are integrated into the ER through the co-translational translocation machinery. We selected EF-Tu receptor (EFR) (Zipfel et al., 2006), brassinosteroid-associated kinase (BAK1)  (Heese et al., 2007), BRI1 (Russinova et al., 2004), and ERECTA receptor (ERec) (Bemis et al., 2013). When coexpressed with StRem1.3 in BiFC complexes, all these RLKs produced stable distribution patterns consistent with ER-PM tethering ( Figure 4B).
Next we tested IMPs that insert into the ER membrane post-translationally. For this purpose we selected tail-anchored (TA) proteins. These proteins lack an N-terminal signal peptide but contain a single transmembrane domain (TMD) which resides so close to the C terminus that it cannot be recognized by the signal recognition particle (SRP) (Rapoport, 2007;Hegde and Keenan, 2011). We selected a set of Arabidopsis TA SNARE proteins, namely SYP21 (Qa-SNARE) (Foresti et al., 2006), VTI11 (Qb-SNARE) (Sanmartín et al., 2007), SYP61 (Qc-SNARE) (Hachez et al., 2014), and VAMP727 (R-SNARE) (Ebine et al., 2008). In these proteins, the C-terminal TMD determines their localization in vesicles of the secretory and endocytic pathways (Supplementary Figure S7). In each case, co-expression of these TA SNARE proteins with StRem1.3 in BiFC complexes resulted in a stable distribution consistent with ER-PM tethering ( Figure 4B).
We also tested several IMPs which span the membrane bilayer more than once, that reside on the PM, endosomal membranes, and vacuolar membrane (Supplementary Figure S7). We tested Arabidopsis DUF679 membrane protein (AtDMP1) (Kasaras et al., 2012), tonoplast potassium channel protein AtTPK1 (Maîtrejean et al., 2011), slow anion channel 1 (SLAC) homolog SLAH3 (Demir et al., 2013), and intracellular aquaporin PIP1 (Wudick et al., 2009). As shown in Figure 4B, a pattern consistent with ER-PM tethering was observed when each of the multipass IMPs was co-expressed with StRem1.3 in BiFC complexes. Together, the above results suggest that patterns consistent with ER-PM tethering were produced with multiple types of IMPs.
Additionally, observations of FLS2, SYP21, or AtDMP1 BiFC complexes formed with StRem1.3 after different times of expression or in cells with different expression levels suggested that the formation of tethering complexes was not an artifact of over-expression (Supplementary Figure S8). While the complexes were barely visible at 18 hpi, the patterns at 36 hpi (the earliest time when visualization was reliable) were almost identical as at 72 hpi. Cells with low levels of BiFC complex formation generally exhibited discrete punctae resembling EPCSs, while in cells with higher levels, the punctae were larger and in some cases merged to form network patterns (Supplementary Figure S8).
In contrast to the IMPs, we did not observe distributions consistent with ER-PM tethering when peripheral membrane proteins were paired with StRem1.3 in BIFC complexes. BIK1-VenusN, PBS1-VenusN, and CPK21-VenusN co-expressed with VenusC-StRem1.3, produced only homogeneously distributed BiFC signals on the PM (Supplementary Figure S9). Similar results were also obtained with Arabidopsis SNAP33 (Kargul et al., 2001;Jahn and Scheller, 2006) which has been recognized as a membrane targeted Qbc-SNARE protein lacking a TMD (Supplementary Figure S9). Collectively, these results imply at least one IMP is required in the BiFC complex to produce ER-PM tethering.

DISCUSSION
In this study, we observed that BiFC complexes containing both the RLK, FLS2, and the membrane-associated remorin protein, StRem1.3, exhibited a range of heterogeneous distribution patterns closely resembling those produced by over-expression of the Arabidopsis ER-PM tether protein, SYT1 (Figures 1, 2). Indeed, co-expression of the FLS2-StRem1.3 BiFC complexes with SYT1 produced fully overlapping distributions (Figure 2), suggesting that the FLS2-StRem1.3 BiFC complexes might act as artificial ER-PM tethering proteins (Figure 6). Since the gap between the ER and the PM is in the 15-20 nm range (McFarlane et al., 2017) and therefore is too small to be resolved by light microscopy (diffraction limit), we used the stability of tethering sites to distinguish them from the more dynamic free ER networks, using kymographs. We note also that since we did not use electron microscopic observations, we cannot rule out that the stable puncta labeled by SYT1-FP are not true contact sites. The co-location of SYT1 with the artificial tethering sites suggests that the artificial sites might develop into true contact sites. In future, co-expressing the tethering constructs with NET3C might provide additional information about the relationship of the tethering sites we observe with bona fide contact sites (Wang et al., 2014).
We showed that StRem1.3 and also the lipid-conjugated peripheral membrane proteins BIK1, PBS1, and CPK21, could replace the PM-binding C2 domain of SYT1 (Figure 3,  Supplementary Figure S6) to produce ER-PM tethering. We showed that the phosphoinositide-binding PH domains from FAPP1 and PLCδ1 could also functionally replace the C2 domain of SYT1 (Figure 5). Finally, we showed that a wide variety of IMPs that transit the ER, including 5 RLKs, 4 tail-anchored proteins, and 4 multi-transmembrane domain proteins, could provide the EM-anchoring function when paired with StRem1.3 as the PM-anchoring protein (Figure 4). In contrast, peripheral membrane proteins that do not transit the ER could not provide the ER-anchoring function (Supplementary Figure S9). On the basis of these observations, we have concluded that FLS2-StRem1.3 BiFC complexes may in fact act as artificial ER-PM tethering proteins. More generally, our model is that any ER-transiting IMP paired with a peripheral membrane protein, either in a BiFC complex or through a direct linkage, may act as an artificial ER-PM tethering protein (Figure 6). In this model, the IMP must transit through the ER, either co-translationally or post-translationally (Walter and Johnson, 1994;Goder and Spiess, 2001). Furthermore, the binding of the peripheral membrane protein to the PM should be sufficiently strong to trap the IMP in the ER, and prevent the completion of the IMP's transit to its final membrane destination. Conversely, the peripheral membrane protein should be synthesized in the cytoplasm and then be targeted to the PM post-translationally, without entering the ER, either via conjugation to a lipid, binding to a PM lipid, or via insertion of a hydrophobic FIGURE 6 | Model for the production of ER-PM tethering complexes via BiFC. Normally, newly synthesized RLK protein FLS2 is targeted to the PM through the ER and transported to the PM via the coat protein complex II (COPII) system. Co-expression of FLS2 and StRem1.3 in BiFC constructs results in rapid spontaneous formation of Venus BiFC complexes tethered to the PM. PM tethering blocks ER-anchored FLS2 from delivery to the PM, resulting in artificial PM-ER complexes. Artificial ER-PM tethering could be created by pairing any ER-transiting integral membrane protein (IMP) with any peripheral membrane protein (PMP) in a BiFC complex.
In plants, several studies have reported observing heterogeneous distributions of BiFC complexes that combine IMPs with peripheral membrane proteins. However, none of these studies have considered the possibility that the distribution patterns observed may have arisen as a result of the formation of artificial ER-PM tethering proteins. For example, Jarsch et al. (2014) showed that whereas the FP-tagged RLK MtNFR1 and remorin MtSYMREM1 were uniformly distributed across the PM when individually expressed in N. benthamiana leaf epidermal cells, when the two were co-expressed in a BiFC complex, the fluorescent signal was exclusively observed in distinct, immobile puncta. Similarly, Demir et al. (2013) observed that BiFC complexes comprised of Arabidopsis SLAH3 (an IMP) and CPK21 (a PMP) localized to distinct PM puncta. Likewise, Bücherl et al. (2017) expressed the following RLK-PMP proteins pairs in BiFC complexes and observed the formation of distinct puncta on the PM: FLS2-BSK1, BRI1-BSK1, FLS2-BIK1, and BRI1-BIK1. Our data suggest that it is necessary to re-evaluate the applicability of BiFC assays for plant membrane studies, and the validity of published studies that used this approach (e.g., Demir et al., 2013;Jarsch et al., 2014;Bücherl et al., 2017) should be re-visited.
Unambiguously determining PM localization in plant cells is challenging. In comparison to mammalian cells, many plant cells contain a large central vacuole that takes up most of the cell volume, resulting in the cytoplasm and organelles being constrained into the periphery of the cell and appressed to the PM. Several methods have been commonly used to distinguish the PM from the vacuolar membrane (tonoplast), including plasmolysis (Speth et al., 2009) and osmolysis (Serna, 2005). However, these methods may be confounded by the presence of the tonoplast or of overexpression artifacts. For example, we observed that some weakly binding PMPs, e.g., SNAP33 (Supplementary Figure S9), SYT1-C2AC2B (Figure 5), and PLCδ1-PH (Figure 5), show substantial cytoplasmic localization when they are over-expressed as FP fusions. The ability of PMPs to form ER-PM tethering complexes may in some circumstances aid in distinguishing between cytosolic and membrane proteins. For example, there is currently not a strong consensus as to the localization of PtdIns(4,5)P 2 in plant cells (Delage et al., 2012). Although PtdIns(4,5)P 2 has been well established as a PM lipid in animal cells (Hammond et al., 2012), evidence for the same localization in plant cells has been ambiguous (Van Leeuwen et al., 2007). Our observations that PLCδ1-PH is effective in forming ER-PM tethering complexes with SYT1n suggests that PtdIns(4,5)P2 is indeed located in the plant PM (but does not rule out other locations as well). In contrast, our negative tethering results with PtdIns(3)P-binding proteins suggest that this lipid does not reside on the cytoplasmic face of the PM.
Given the ability of IMPs to act as the ER-anchor in artificial ER-PM tethering complexes, such complexes could possibly be used to investigate whether a protein may be an IMP or not, as suggested previously by Zamyatnin et al. (2006). Bioinformatic analysis has been increasingly used to predict the identity and topology of IMPs. However, these algorithms are not fully accurate. For example, two commonly used programs TMHMM (Krogh et al., 2001) and Protter (Omasits et al., 2014) can differ in their predictions. Artificial ER-PM tethering complexes could be used to test such predictions. As one example from this work, Protter and TMHMM both predicted a weak TMD in CPK21 whereas all other members of the CDPK family are targeted to the PM by myristoylation and palmitoylation (Speth et al., 2009;Asai et al., 2013;Schulz et al., 2013), suggesting that the bioinformatic prediction may be incorrect. In fact, we observed that CPK21-SYT1n BiFC complexes exhibited tethering (Supplementary Figure S6) but CPK21-StRem1.3 BiFC complexes did not (Supplementary Figure S9), confirming CPK21 as a PMP, not an IMP. Further examples will be needed to determine if this approach is generally useful.
Genetically designed chimeric proteins have been successfully developed to manipulate tethering of the ER to the PM or to other organelles, and to study cellular processes involving tethering proteins (Kornmann et al., 2009;Chang et al., 2013;Bockler and Westermann, 2014;Poteser et al., 2016;Lee et al., 2019). For example, a chimeric protein named MAPPER was used as a constitutive ER-PM tethering marker to investigate the molecular mechanism for dynamic regulation of ER-PM tethering during Ca 2+ signaling in live mammalian cells; MAPPER is derived from the human ER-PM tether STIM1 and contains minimal ER and PM-targeting motifs, linked by a fluorescent protein (Chang et al., 2013;Poteser et al., 2016). More recently, MAPPER has also been used as a non-regulated ER-PM tethering marker to study the response of Arabidopsis SYT1 on the regulation of ER-PM connectivity under ionic stress (Lee et al., 2019). Additionally, ChiMERA, a synthetic ER-Mitochondria tether consisting of GFP fused to the mitochondrial membrane anchored TMD motif of Tom70 and an ER tail-anchor motif from Ubc6, was used to restore mitochondria-ER contacts in yeast mutants (Kornmann et al., 2009;Bockler and Westermann, 2014). Therefore, our work here not only suggests a potential way to develop fluorescent molecular markers for EPCSs, but also suggests a molecular tool to manipulate tethering of the ER to the PM, or even to other membrane organelles in plants. One can imagine, for example, tethering complexes in which dimerization of the two components is regulated by small molecules and/or light as reported previously (Karginov et al., 2011;Guntas et al., 2015).
In summary, we have deployed an extensive toolset of plant membrane marker proteins and mutant controls (summary in Table 1, Supplementary Table S1) to characterize the artificial ER-PM tethering that may result from spontaneous reassembly of fragmented fluorescent proteins during co-localization studies. These results complement our findings that a similar phenomenon can produce tethering of multi-vesicular bodies and the tonoplast to the PM (Tao et al., 2019). Our results indicate the possibility of new tools for deliberately manipulating ER-PM tethering, while at the same time highlighting a previously unrecognized artifact that may have confounded several published studies.

Plant Materials
Nicotiana benthamiana and A. thaliana plants were grown in soil (Fafard R 4M Mix). N. benthamiana plants were grown in a growth chamber with a 14 h photoperiod at 25 • C for 5 weeks before A. tumefaciens infiltration assays. A. thaliana seeds were sown in soil and left at 4 • C for 3 day of cold stratification. Then the seedlings were grown in a growth chamber with a 12 h photoperiod at 20 • C for 4 weeks before protoplast isolation.

Transient Expression in N. benthamiana Leaves and A. thaliana Protoplasts
The procedures to introduce expression vectors into A. tumefaciens strain GV3101, and to infiltrate transformed A. tumefaciens cells into 5-week-old N. benthamiana leaves were carried out as described previously (Lu et al., 2013). A. tumefaciens cells were infiltrated at OD 600 of 0.1 for the expression of the full-length fluorescent protein tagged proteins; for co-expression of BiFC constructs, two A. tumefaciens cultures with OD 600 of 0.2, respectively, were equally mixed together to reach the final OD 600 at 0.1. All infiltrated A. tumefaciens cells were suspended in MES buffer (10 mM MgCl 2 , 10 mM MES pH 5.7, and 100 µM acetosyringone). N. benthamiana leaves were imaged at 3 days post infiltration. A. thaliana mesophyll protoplasts were prepared from leaves of 4-week-old seedlings, and 10 µg of plasmid DNA were used for each transformation as described (Yoo et al., 2007). Following transformation, protoplasts were suspended overnight in W5 buffer (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl, 2 mM MES pH 5.7) at 25 • C before observation.  described (Günl et al., 2011). All microscopy images were obtained using a ZEISS LSM 780 NLO confocal microscope system equipped with a 458-nm argon laser for CFP (emission wavelength 560-509 nm), a 514-nm argon laser for YFP/Venus (emission wavelength 518-553 nm), and a 561-nm Diode Pumped Solid State (DPSS) laser for tagRFP and FM4-64 (emission wavelength 562-640 nm). For time-lapse imaging, 100 consecutive frames without time intervals (combined speed of about 0.78 fps) were acquired sequentially. The kymograph plots were generated using ImageJ (Version 1.51n, NIH) and the plug-in "KymographBuilder" with a 30 µm segmented line for this measurement. When interpreting the kymographs, close attention was paid to distinguishing the cortical ER from transvacuolar strands, which are also dynamic. All microscopy images were processed using the Zeiss ZEN2 (Blue edition) program.

AUTHOR CONTRIBUTIONS
KT and BT conceived and planned the experiments and wrote the manuscript. KT performed experiments and analyzed the data. JW and FA were involved in performing experiments.

ACKNOWLEDGMENTS
We thank Anne-Marie Girard for technical assistance, Prof. Libo Shan (Texas A&M) for providing several plasmid constructs for sub-cloning. This research was supported by Oregon State University.