Signal motif-dependent ER export of the Qc-SNARE BET12 interacts with MEMB12 and affects PR1 trafficking in Arabidopsis

Compartmentalization of cells in eukaryotes presupposes the development of mechanisms for protein trafficking between different membrane-enclosed organelles. Vesicular transport is the predominant pathway for protein trafficking in eukaryotic cells. Multiple molecular machineries are required for the formation, transport, tethering and fusion of the vesicles to the target compartment (Bonifacino and Glick, 2004). Soluble N-ethylmaleimide-sensitive fusion protein attachment protein receptors (SNAREs) have been identified as critical components involved in the final fusion step for the vesicular transport pathways (Söllner et al., 1993). These facilitate vesicle–target-membrane fusion by forming hetero-tetrameric trans-SNARE complexes derived from a specific set of SNARE proteins (Jahn and Scheller, 2006; McNew et al., 2000).

A large number of SNARE proteins are encoded in the plant genome (Sanderfoot, 2007; Sanderfoot et al., 2000). Numerous studies have unraveled the important role of SNARE proteins in plants, involving various biological processes including pathogen defense, cytokinesis, abiotic stress, cell expansion, symbiosis, gravitropism, gametophyte and seed development (Ebine et al., 2008; El-Kasmi et al., 2011; Grefen et al., 2010a; Hachez et al., 2014; Honsbein et al., 2009; Huisman et al., 2016; Pan et al., 2016; Reichardt et al., 2007; Uemura et al., 2012b; Yano et al., 2003). The precise targeting of SNARE proteins to a distinct compartment is essential for mediating the vesicle–target-membrane fusions which secures an efficient and accurate protein trafficking. Mis-targeting of SNARE proteins results in numerous cellular defects. For instance, the cell plate formation is disrupted in a mutant with an impaired trafficking of the syntaxin KNOLLE (Park et al., 2013; Teh et al., 2013). In addition, a recent study has revealed the novel role of the endoplasmic reticulum (ER)-associated SNARE SYP73 in maintaining the ER integrity and consequently streaming, since these features were altered in a syp73 mutant (Cao et al., 2016). Therefore, correct localization of SNARE proteins seems to be a prerequisite for their proper functioning in different cellular processes.

Most SNARE proteins are tail-anchored (TA) proteins, their membrane association being conferred by the C-terminal transmembrane domain (TMD). TA proteins are post-translationally inserted into the membrane through the guided entry of TA proteins (GET) pathway (Schuldiner et al., 2008; Stefanovic and Hegde, 2007). Recent studies have demonstrated the functional GET components for TA protein targeting in Arabidopsis (Xing et al., 2017) for the process whereby the SNARE SYP72 is inserted into the ER membrane through the GET pathway (Srivistava et al., 2016). Once translocated into the ER, further targeting of membrane proteins is determined by either specific signal motifs, the length of TMD or a combination of both (Brandizzi et al., 2002; Hanton et al., 2006, 2005b; Matheson et al., 2006; Rojo and Denecke, 2008; Saint-Jore-Dupas et al., 2006). In the early secretory pathway of plants, proteins that are exported from the ER and traffic to the Golgi are mediated by coat protein complex II (COPII) machinery (Brandizzi and Barlowe, 2013; DaSilva et al., 2004; Hawes et al., 2008; Moreau et al., 2007; Robinson et al., 2015; Stefano et al., 2014). According to the model in yeast and mammals, the formation of COPII vesicles is initiated by the recruitment of the small GTPase SAR1 to the ER membrane. Activated SAR1 then recruits the inner-coat dimeric complex SEC23–SEC24 which captures protein cargos. Subsequent recruitment of the outer-coat complex SEC13–SEC31 stimulates the GTP hydrolysis of activated SAR1, and eventually leads to the formation of COPII carriers containing protein cargos to be exported (Bassham et al., 2008; Chung et al., 2016; Hwang and Robinson, 2009; Marti et al., 2010). Among the Golgi-localized SNARE proteins (Uemura et al., 2004), SYP31 and MEMB11 have been reported to have a critical role in mediating ER-to-Golgi anterograde transport (Bubeck et al., 2008; Chatre et al., 2005). Although the roles of SNARE proteins in the early secretory pathway have been investigated, the mechanism for SNARE targeting to the Golgi membrane remains elusive. Until now, only a single study has demonstrated that ER export and Golgi targeting of SYP31 depends on the di-acidic motif in its N-terminus (Chatre et al., 2009).

In the present study, we aimed to elucidate the targeting mechanism and the functional role of the Qc-SNARE BET12 (also termed Bs14b) in the early secretory pathway. Previous studies suggested that BET12 is involved in plant fertility and displayed a Golgi localization in Arabidopsis protoplasts (Bolanos-Villegas et al., 2015). BET11 (also termed Bs14a), shares a 78% amino acid similarity to BET12, and was also suggested to localize on Golgi membranes (Uemura et al., 2004). Unlike SYP31 and MEMB11, BET11 overexpression did not severely affect protein ER-to-Golgi anterograde transport (Bubeck et al., 2008; Chatre et al., 2005). Strikingly, an in vitro study in yeast has suggested that BET11 and BET12 tend to form a distinct quaternary SNARE complex with different yeast Golgi SNAREs, as BET11 had a SNARE-binding profile that resembled that of Sft1p, whereas the binding profile of BET12 resembled that of Bet1p (Tai and Banfield, 2001). In order to characterize the role of BET12 in ER-to-Golgi protein trafficking, we first determined the subcellular localization of BET12 in transgenic plants and uncovered a signal motif-dependent targeting mechanism that was necessary for efficient ER export. We further identified the Qb-SNARE MEMB12 as the interacting partner of BET12 and performed experiments that suggested they have a function in regulating the secretion of pathogenesis-related protein 1 (PR1; At2g14610) in Arabidopsis.

To determine the subcellular localization of BET12, we generated transgenic plants expressing N-terminally yellow fluorescent protein (YFP)-tagged BET12 (YFP–BET12) driven by the UBQ10 promoter. Confocal laser-scanning microscopy (CLSM) analysis revealed a punctate distribution pattern of YFP–BET12 in Arabidopsis root cells. Immunofluorescence labeling using antibodies against organelle-specific markers was performed, and revealed that YFP–BET12 was partially colocalized with the Golgi marker EMP12 (also known as TMN1) and the trans-Golgi network (TGN) marker SYP61 but was distinct from the ER marker calreticulin (Fig. 1A–C), suggesting that YFP–BET12 localizes to both the Golgi and TGN.

Previous studies have reported that the fungal toxin Brefeldin A (BFA) causes aggregation of both Golgi and TGN, with the Golgi-derived and TGN-derived aggregates being distinct from each other (Lam et al., 2009; Zhang et al., 2011a). To further prove the localization of YFP–BET12, we carried out BFA treatment followed by a styryl dye FM4-64 uptake experiment, since the dye can be used as an endocytic tracer to label endosomal compartments, including the TGN (Bolte et al., 2004), using transgenic plants expressing ST–YFP (a trans-Golgi marker; ST is the rat sialyl transferase ST6GAL1), VHAa1–GFP (TGN marker) and YFP–BET12. After BFA treatment, VHAa1–GFP was found in dense-core aggregates, termed BFA bodies, together with FM4-64, while ST–YFP was distinctly found at the periphery of the FM4-64-labeled BFA body (Fig. S1A,B). Line plots were constructed to show the corresponding fluorescence intensity along the BFA-induced aggregates. From these, the VHAa1–GFP peak was seen to completely overlap with the FM4-64 peak, whereas the ST–YFP peak was largely separate from the FM4-64 peak (Fig. S1A,B). Similarly, the fluorescence pattern of YFP–BET12 changed from punctate to aggregates upon BFA treatment (Fig. 1D). However, unlike ST–YFP and VHAa1–GFP, the YFP–BET12 aggregate was not only found in the dense-core FM4-64 aggregate but also at its periphery (Fig. 1D). Fluorescence intensity line plots along the YFP–BET12 aggregate showed a different distribution pattern from ST–YFP and VHAa1–GFP in that the YFP–BET12 signal intensity remained high in both the periphery and the core region of the FM4-64 peak (Fig. 1E), indicating that YFP–BET12 is being trapped in both the Golgi-derived and TGN-derived aggregates.

To further confirm the localization of YFP–BET12 from the CLSM analysis, we performed immunogold electron microscopy (EM) with anti-GFP antibodies (recognizing YFP) on ultrathin sections prepared from high-pressure frozen and freeze-substituted root cells of transgenic Arabidopsis seedlings expressing YFP–BET12. Consistent with our confocal findings, EM observations showed that gold particles (on secondary antibodies recognizing anti-GFP) were present on both the Golgi stacks as well as the associated TGN (Fig. 1F). Quantification of immunogold labeling indicated that ∼37% and ∼52% of the gold particles were associated with the Golgi (including on both the cis- and trans-side) and TGN, respectively (Fig. 1G). Taken together, CLSM and EM studies demonstrated that YFP–BET12 localized to both the Golgi and TGN in transgenic Arabidopsis plants.

In order to elucidate the BET12-targeting mechanism, we first needed to know its protein topology. Most SNARE proteins are type II membrane protein, with the N-terminus that contains the SNARE motif exposed to the cytosol and with a C-terminal TMD. To determine whether BET12 is a membrane protein, total proteins were extracted from Arabidopsis cells expressing YFP–BET12 and were then separated into soluble and membrane fractions by ultracentrifugation. Immunoblot analysis using anti-GFP antibodies showed that YFP–BET12 was found in the membrane fraction, but not the soluble fraction (Fig. S2A, lane 1 and 2). To distinguish integral from peripheral membrane protein, microsomes isolated were then subjected to high salt, high pH and detergent washes, followed by immunoblotting with anti-GFP antibodies. The reliability of the assay was verified by using anti-VSR antibody (this antibody recognizes several VSR proteins), as VSR is known to be an integral membrane protein (Paris et al., 1997). Immunoblot analysis showed that YFP–BET12 remained associated with microsomal membrane under high salt and high pH condition but released to the soluble fraction upon detergent washes (Fig. S2A, lanes 3–10). The response of YFP–BET12 towards different conditions was similar to that of VSR, indicating that BET12 is most likely to be an integral membrane protein.

As determined by sequence analysis (TMHMM server 2.0), BET12 is predicted to have a typical SNARE topology (Fig. S2B). To determine whether the YFP–BET12 fusion retains the same topology, we carried out a protease protection assay using microsomes isolated from Arabidopsis cells expressing YFP–BET12, followed by immunoblotting with anti-GFP antibodies. GFP–VSR2, a membrane protein with a known topology that the N-terminus is facing into the lumen (Cai et al., 2011), was used as a control in this assay. No band was detected in the immunoblot analysis of YFP–BET12 in the presence of the protease trypsin, indicating that the N-terminus together with the YFP face the cytosol and can therefore be digested by trypsin (Fig. S2C, lane 2). By contrast, a band was detected in the immunoblot of GFP–VSR2, showing that its lumen-facing N-terminus was protected from trypsin digestion (Fig. S2C, lane 5). As expected, no band could be detected for YFP–BET12 and GFP–VSR2 in the presence of both trypsin and Triton X-100 (Fig. S2C, lane 3 and 6). The results from these biochemical assays suggest that YFP–BET12 is an integral membrane protein, which maintains a typical SNARE topology with its N-terminus facing the cytosol.

With the determined BET12 protein topology, we next generated various truncation and deletion versions of YFP–BET12, followed by transient expression in Arabidopsis protoplasts to identify the regions that are responsible for BET12 trafficking. Upon transient expression of YFP–BET12 and CLSM analysis with organelle-specific markers, it could be seen that YFP–BET12 partially colocalized with the Golgi marker Man1–RFP and TGN marker mRFP–SYP61 (Fig. 2A,B). As YFP–BET12 displayed a partial Golgi and TGN localization in both the transiently expressing protoplasts and the transgenic plants, we decided to use the transient setup to screen for targeting defects caused by truncation and deletion of YFP–BET12. We first deleted the whole cytosolic N-terminus of BET12 to determine its effect on trafficking. CLSM analysis showed that YFP–BET12(107-130) colocalized with the ER marker CNX–RFP and displayed an ER pattern (Fig. 2C), indicating that the presence of the TMD and C-terminus is not sufficient for its ER export, and the cytosolic N-terminal probably contains ER export signals. Since, in previous studies, the SNARE motif has been suggested to play a role in SNARE targeting (Joglekar et al., 2003), we also deleted the BET12 SNARE motif to test for any targeting defect. Upon transient expression, YFP–BET12(1-32)(98-130) showed a punctate pattern which colocalized with Man1–RFP (Fig. 2D), suggesting that SNARE motif is not essential for BET12 ER export. To identify the region containing the ER export signals, we further deleted the nine amino acids (a.a.) between the SNARE motif and the TMD. Interestingly, instead of only showing a punctate pattern, YFP-BET12(1-32)(107-130) partially colocalized with CNX–RFP and displayed an ER pattern as well (Fig. 2E). This result suggests that the very N-terminus (a.a. 1–32) contains ER export signal and aids in targeting to the Golgi which lead to the punctate pattern (Fig. 2F). The deletion of the nine amino acids (a.a. 98–106) inhibits ER export to some extent, thus leading to the appearance of the ER pattern. Similarly, the presence of both the punctate and ER pattern was also observed when expressing YFP–BET12(98-130) (Fig. 2G,H), supporting the notion that the nine amino acids (a.a. 98–106) also contain an ER export signal and aid in BET12 targeting, while the absence of the very N-terminus (a.a. 1–32) hampers its ER export efficiency. Taken together, in vivo expression of truncated and deleted versions of YFP–BET12 revealed that both the N-terminus (a.a. 1–32) and the region between the SNARE motif and TMD (a.a. 98–106) of BET12 contain ER export signals and are responsible for its efficient trafficking.

To elucidate the mechanism of BET12 trafficking, we made use of the region identified as containing the ER export signals to find out the potential interaction partners and the related trafficking machinery. A synthetic peptide of the nine amino acids (a.a. 98–106 containing the ER export motif) was conjugated onto Sepharose beads as bait for pulldown experiments. Sepharose beads conjugated with or without this peptide were incubated with total proteins extracted from Arabidopsis suspension cells. After washing, proteins were eluted from the beads and subjected to SDS-PAGE followed by silver staining (Fig. 3A). In duplicate experiments, intense bands that were repeatedly observed in the lane with the conjugated peptide but not with the empty Sepharose were isolated for tandem mass spectrometry (MS/MS) analysis. MS/MS analysis showed that certain proteins pulled out by the nine-amino-acid peptide were identified as components of the COPI (ε-COP and ζ-COP) and COPII (Sar1) machineries (Table S1). It has been reported that COPI and COPII machinery are important for regulating ER-to-Golgi protein trafficking (Gao et al., 2014; Hanton et al., 2005a; Paul and Frigerio, 2007). As a Golgi-localized SNARE, BET12 likely interacts with the COPI and COPII machinery components to maintain its proper localization. To determine whether this was indeed the case, we transiently expressed YFP–BET12 together with a GDP-fixed mutant form of ADP-ribosylation factor 1 (Arf1-GDP), which interferes with the COPI machinery (Pimpl et al., 2003; Takeuchi et al., 2002), and with the GTP-locked version of SAR1 (SAR1-GTP) mutant, which interferes with the COPII machinery and inhibits protein export from the ER (Osterrieder et al., 2010; Takeuchi et al., 2000; Zeng et al., 2015). CLSM analysis showed that YFP–BET12 relocated to the ER and colocalized with the ER marker CNX–RFP upon co-expression with Arf1-GDP (Fig. 3B). Similarly, YFP–BET12 displayed an ER pattern when co-expressed with the SAR1 GTP-locked mutant form Sar1C-DN–RFP, indicating its failure to undergo ER export when the COPII machinery is disrupted (Fig. 3C). Pulldown MS/MS analysis, together with the evidence obtained from the in vivo cell biological study, suggested that BET12 trafficking depends on functional COPI and COPII machineries.

The intimate relationship of BET12 with COPII machinery prompted us to take a detailed look into the N-terminus (a.a. 1–32) and the region between the SNARE motif and TMD (a.a. 98–106) in order to identify more precisely amino acid residues responsible for COPII-mediated ER export. By sequence alignment, we identified motifs that bind and interact with Sar1 and Sec24, as reported previously. In yeast, it has been shown that the COPII inner coat complex Sec23–Sec24 binds to LxxLE and mediates SNARE Bet1 ER export (Mossessova et al., 2003). Interestingly, the COPII-binding motif LxxLE was present within the N-terminus region (a.a. 1–32) of BET12 (Fig. 4A, in blue). In addition, the dibasic motif of glycosyltransferase was found to interact with Sar1 in mammalian cells (Giraudo and Maccioni, 2003). A similar dibasic motif was identified in the region between the BET12 SNARE motif and the TMD (a.a. 98–106) (Fig. 4A, in blue). To prove the conserved role of these motifs in ER export, we generated two mutated BET12 constructs with point mutations in the LxxLE and RK dibasic motif respectively, which are termed YFP–BET12(L18A,L21A,E22A) and YFP–BET12(R102A,K103A). We then performed a transient expression experiment using Arabidopsis protoplasts, and the subcellular localization of these mutated BET12 proteins was determined by CLSM. Unlike the sole punctate pattern displayed by YFP–BET12 (Fig. 4B), both the YFP-BET12(L18A,L21A,E22A) and YFP-BET12(R102A,K103A) showed a punctate and ER pattern which partially colocalized with CNX–RFP (Fig. S3A,B), suggesting that the separate mutations in one of the potential COPII-binding motifs causes a partial defect in BET12 ER export. To further assess the role of these motifs, we generated a construct with five point mutations so that all five residues became alanine residues, termed YFP–BET12-m (Fig. 4A, in red). Strikingly, YFP–BET12-m displayed a dominant ER pattern and colocalized with CNX–RFP (Fig. 4C), suggesting that the mutations severely inhibit its ER export. Interestingly, a few puncta with faint fluorescence signals were localized close to the ER, suggesting that a limited amount of YFP–BET12-m may still exit the ER and reach the Golgi (Fig. 4C; Fig. S4). The failure in ER export of YFP–BET12-m was not only observed in protoplasts but also in intact Arabidopsis seedlings; YFP–BET12 and YFP–BET12-m was expressed in 7-day-old seedlings by particle bombardment and the corresponding transformed cells were imaged using CLSM. In leaf pavement and trichome cells, YFP–BET12 displayed a punctate pattern resembling the Golgi localization (Fig. 4D). By contrast, YFP–BET12-m exhibited an ER network pattern in both the leaf pavement and trichome cells, which was obviously observed throughout the z-stack projection of multiple confocal layers (Fig. 4E). Point mutagenesis of the five residues significantly changed the localization of YFP–BET12 from a punctate to an ER pattern. The fact that there are two putative COPII-binding motifs in different regions of the protein may explain our above observation that the absence of any one ER export signal leads to both the punctate and ER localization pattern.

The role of the SNARE BET12 in ER–Golgi trafficking is unclear. We speculated that the overexpression of an ER-export-defective form of BET12 might interfere with protein trafficking in the early secretory pathway. To test this hypothesis, we co-expressed YFP–BET12-m with protein cargos known to be transported through the conventional secretory pathway. Aleurain–mRFP and RFP–SCAMP1 were used as soluble and membrane cargo markers, respectively (Lam et al., 2007; Miao et al., 2008). However, CLSM analysis showed that the trafficking of both the Aleurain–mRFP and RFP–SCAMP1 was not affected by YFP–BET12-m and they were transported to the vacuole and plasma membrane correspondingly (Fig. S4A,B). Man1–RFP still maintained a typical punctate pattern upon co-expression with YFP–BET12-m (Fig. S4C), suggesting the Golgi was not disrupted. We then screened a set of Golgi-localized SNARE proteins for any trafficking defect as YFP–BET12-m may interact with its partner SNARE proteins and are retained together in the ER. Interestingly, CLSM analysis revealed that only mCherry–MEMB12 but not mCherry–SYP31 nor mCherry–GOS12 was trapped in the ER together with YFP–BET12-m in protoplasts (Fig. 5A; Fig. S4D,E). The ER-trapping effect of YFP–BET12-m on mCherry–MEMB12 was also observed in intact Arabidopsis leaf and trichome cells (Fig. 5B,C). Consistent with the previous findings, mCherry–MEMB12 displayed a punctate pattern when co-expressed with YFP–BET12 (Fig. S5). The ER trapping of mCherry–MEMB12 is probably caused by the physical interaction with the ER-export-defective YFP–BET12-m. To confirm this, we performed a fluorescence resonance energy transfer-acceptor photobleaching (FRET-AB) assay to verify the potential protein–protein interaction between YFP–BET12-m and Cerulean–MEMB12 (Xing et al., 2016) in which YFP–linker–Cerulean, and YFP–BET12-m with CNX–Cerulean was used as a positive and negative control, respectively. FRET-AB analysis suggested an in vivo interaction between YFP–BET12-m and Cerulean–MEMB12 as their FRET efficiency was significantly high compared to the negative control (Fig. 5D). Co-immunoprecipitation (Co-IP) assay using GFP-trap was further performed to show the interaction between YFP–BET12-m and HA–MEMB12, as HA–MEMB12 was immunoprecipitated by YFP–BET12-m as shown in the immunoblot using anti-GFP and -HA antibodies (Fig. 5E). Taken together, results from the CLSM analysis and protein–protein interaction assays strongly suggest that YFP–BET12-m interacts with MEMB12 and its ER-export-defective nature causes the trapping of both proteins in the ER.

A previous study suggested that MEMB12 is involved in regulating the secretion of the antimicrobial protein PR1 in Arabidopsis (Zhang et al., 2011b). As BET12 was shown to interact with MEMB12, we decided to test whether BET12 would also affect PR1 secretion. Upon transient expression of PR1–RFP in Arabidopsis protoplasts, we could not detect any intracellular fluorescence signal using confocal microscopy. We speculated that the majority of PR1–RFP was being secreted to the culture medium. To prove this hypothesis, we therefore separated the protoplasts from the culture medium by low-speed centrifugation. The collected culture medium was then concentrated and total proteins were extracted from the protoplast pellet. Both the medium and protein extracts were then subjected to immunoblot analysis using anti-RFP antibodies. As expected, a band was detected in the medium fraction but not in the proteins extracted from pellet (Fig. 6A, lane 1, 2), suggesting that PR1–RFP is constantly secreted out of the protoplasts. Interestingly, co-expression of PR1–RFP with YFP–BET12 or YFP–MEMB12 resulted in intracellular accumulation of PR1–RFP, as evidenced by the band detection in the pellet fraction (Fig. 6A, lane 4, 6), although the majority of PR1–RFP was still secreted to the medium (Fig. 6A, lane 3, 5). The amounts of PR1–RFP retained intracellularly and secreted extracellularly were quantified (Fig. 6B), showing that the ectopic expression of YFP–BET12 and YFP–MEMB12 affects PR1 trafficking. CLSM analysis using Arabidopsis seedlings expressing PR1–RFP was performed to further confirm its secretion behavior. When PR1–RFP was expressed alone, PR1–RFP showed a fluorescence signal in the extracellular space surrounding the leaf pavement cell (Fig. 6C). Consistent with this, co-expression of PR1–RFP with YFP–BET12 affected PR1–RFP trafficking as red fluorescence foci were found inside the cell while some PR1–RFP was still secreted to the apoplast (Fig. 6D). To further verify the relationship between BET12 overexpression and PR1 trafficking, we repeated the secretion assay with a gradual increment in the expression of HA–BET12 and determined its effect in PR1 trafficking by performing immunoblot analysis. The greater the abundance of HA–BET12 is, the more PR1–RFP was trapped in the pellet fraction (Fig. 6E), suggesting that the effect of HA–BET12 in PR1 intracellular accumulation is dosage dependent. Interestingly, overexpression of HA–BET12-m also interfered with PR1 secretion and caused its intracellular retention, although to a lesser extent than HA–BET12 did (Fig. S6). Taken together, both the biochemical assay and in vivo imaging data suggest that ectopic expression of BET12 affects PR1 trafficking and causes its intracellular accumulation.

PR1 is known as an antimicrobial protein that plays an important role in plant immunity (Van Loon and Van Strien, 1999). As ectopic expression of YFP–BET12 interferes with PR1 trafficking, to test whether the antibacterial defense was affected, we performed bacterial growth assays using wild-type and transgenic plants overexpressing YFP–BET12 infected by both virulent and avirulent (avrRpt2) strains of the bacterial pathogen Pseudomonas syringae pv. tomato (Pst DC3000). Bacterial growth assays showed that there was no significant difference in the growth of the virulent or avirulent Pst strains when comparing the wild-type and YFP–BET12 overexpression plants (Fig. 6F), suggesting that YFP–BET12 transgenic plants are not more susceptible to pathogen infection.

Multiple mechanisms for membrane protein targeting have been proposed in plants, including distinct signal motif recognition by trafficking machineries, as well as the various properties of the TMD (Brandizzi et al., 2002; Langhans et al., 2008; Robinson et al., 2007; Rojo and Denecke, 2008; Schoberer et al., 2009; Wang et al., 2014). However, the targeting mechanisms for most of the SNAREs to their destined compartments, remain obscure in plants. It has been reported that the entire longin domains of the VAMP7 SNAREs and the di-acidic motif of the SYP31 are essential for the vacuolar and Golgi targeting, respectively (Chatre et al., 2009; Uemura et al., 2005). However, additional factors and the nature of the trafficking machinery involved in SNARE targeting is not known. In this study, we demonstrated the Golgi and TGN localization of the Qc-SNARE BET12 in Arabidopsis and revealed its COPII-dependent ER export mechanism. Interestingly, subcellular localization studies of BET12 using transgenic plants and protoplasts yielded results with minor discrepancy: YFP–BET12 was more highly localized to the TGN in transgenic plant root cells while YFP–BET12 showed a slightly higher colocalization rate with the Golgi than the TGN marker in protoplasts. This variation may be due to the different methodologies (immunolabeling and transient expression) and plant materials (transgenic plants and protoplasts) used for the localization studies. Previous studies indicated that the TMD length is unlikely to be the sole determinant that dictates subcellular localization of SNARE proteins (Chatre et al., 2009; Uemura et al., 2005). This is consistent with our findings showing that the presence of the TMD of BET12 is alone insufficient for its ER export (Fig. 2). It has been reported that the mammalian ortholog rat BET1 (rBET1) depends on its SNARE motif for targeting (Joglekar et al., 2003). However, the absence of the BET12 SNARE motif did not affect its ER export in Arabidopsis cells. Instead, through truncation experiments, we identified two regions that contain signal motifs that are responsible for efficient ER export of BET12: the N-terminus (a.a. 1–32) and the region in-between the SNARE motif and the TMD (a.a. 98–106). BET12 ER export is hampered but not completely abrogated by deleting either one of the signal-motif-containing regions (either a.a. 1–32 or a.a. 98–106), as evidenced by the observation of both the ER and punctate (Golgi) localization pattern of the truncated YFP–BET12 (Fig. 2). Indeed, similar experimental approaches have been applied in elucidating the targeting mechanism of another Golgi-localized SNARE SYP31. A novel di-acidic motif ExxD, residing in a region between the SNARE helices, was found to facilitate the ER export of SYP31 (Chatre et al., 2009). To precisely identify the amino acid residues constituting the signal motifs for BET12 ER export and the trafficking machinery involved, mutagenesis and pulldown experiments were performed. Synthetic peptide pulldown experiments indicated that Sar1, one of the major constituents for COPII vesicle formation, may interact with the region between the SNARE motif and the TMD (a.a. 98-106) (Fig. 3). Consistent with this, by sequence analysis, we found that the presence of a dibasic motif, which is reported to be Sar1 binding (Giraudo and Maccioni, 2003; Srivastava et al., 2012; Yuasa et al., 2005), in the same region. It has been shown that mutation of the dibasic motif residues proximal to the TMD of bZIP28 to alanine residues interferes with its interaction with Sar1 and inhibits its ER export under ER stress (Srivastava et al., 2012). In addition, a conserved motif, LxxLE, in yeast Bet1 which binds to the B site of the Sec23–Sec24 complex (Mossessova et al., 2003), was identified in the BET12 N-terminal region (a.a. 1–32). Although no direct interaction between the SNAREs and Sec24 has been reported in plants, Sec24 has been shown to interact with the potassium channel KAT1 via its di-acidic motif and facilitate its ER export (Sieben et al., 2008). Independent mutagenesis of the putative Sec24-binding motif LxxLE (a.a. 18–22) and the Sar1-binding dibasic motif RK (a.a. 102–103) into alanine residues results in a partial defect in ER export of BET12 (Fig. S3). Simultaneous mutagenesis of all the five residues severely inhibited BET12 trafficking, which resulted in the ER localization of YFP–BET12-m (Fig. 4), as well as in an ER-trapping effect upon co-expression with Sar1C-DN–RFP. These data indicate that the ER export of BET12 is signal motif-dependent and is mediated by functional COPII machinery.

In addition to COPII, the COPI machinery component Arf1 has been reported to interact with another Golgi-localized SNARE, MEMB11 (Marais et al., 2015). It has been proposed that membrin (also known as GOSR2), the mammalian MEMB11, act as a recruiter for Arf1 recruitment to the Golgi membrane to initiate COPI vesicle formation (Honda et al., 2005). EMP12 was also shown to interact with the COPI machinery to maintain its Golgi localization (Gao et al., 2012). Interestingly, certain COPI machinery components were identified in our pulldown MS/MS data (Fig. 3), suggesting that BET12 may also bind to COPI and become incorporated into COPI vesicles for its proper targeting.

Impaired trafficking and mis-targeting of SNARE proteins has been reported to be detrimental to cells. For instance, the correct trafficking of KNOLLE is important for plant cell cytokinesis (Park et al., 2013; Reichardt et al., 2007), while a reduced salt tolerance may be caused by the partial mislocalization of SYP61 in tno1 mutants (Kim and Bassham, 2011). To determine whether there is any adverse cellular effect caused by the ER-export-defective form of BET12, we monitored the trafficking of both the soluble and membrane proteins known to employ the ER-to-Golgi secretory pathway. However, no trafficking defect was observed as both the Aleurain–mRFP and RFP–SCAMP1 were transported to their respective destined compartments and even YFP–BET12-m was significantly trapped in the ER (Fig. S4). Interestingly, it is reported that the overexpression of the ER-export-defective form of SYP31 severely inhibits tobacco plant growth (Melser et al., 2009), as the accumulation of ER-trapped SYP31 is toxic to the secretory pathway by potentially disturbing the homeostasis of the SNARE machinery as well as the ER–Golgi interface. Although protein trafficking was not affected by YFP–BET12-m overexpression, the defective ER export of BET12 could trap the interacting SNARE partner together in the ER. Among the Golgi-localized SNAREs screened, only MEMB12 was found to be trapped in the ER upon co-expressing with YFP–BET12-m (Fig. 5). Bos1, the yeast homolog of MEMB12, does not have any Sec24-binding motif (Mossessova et al., 2003). Unlike other Golgi-localized SNAREs that can bind to Sec24 for their ER export, it is plausible that Bos1 binds to its partner SNAREs to form a complex, which is then co-packaged into COPII vesicles for ER export (Mossessova et al., 2003). Similar to its ortholog Bos1, no putative Sec24-binding motif could be identified in MEMB12, indicating that its ER export may be dependent on its partner SNARE. In this sense, the ER retention of YFP–BET12-m, which interacts with MEMB12, may result in the defective ER export of MEMB12 as well. Interestingly, a recent study suggested that the preassembled ternary Golgi Q-SNARE complex was preferred for ER export in mammalian cells, a finding that is supported by the sorting defect present in all other Q-SNAREs upon the mutation of the binding site for Sec24C–Syntaxin 5 (Adolf et al., 2016). In our study, the pulldown assay followed by the MS/MS analysis failed to identify the potential SNARE partners interacting with BET12, probably due to the fact that only the defined region (a.a. 98–106) of BET12 but not its SNARE coiled-coil domain was used as a bait for the pulldown assay. Indeed, BET12 was shown to preferentially form a SNARE complex with the yeast Bos1 and certain Golgi SNAREs in an in vitro study (Tai and Banfield, 2001). Similar in vitro binding assays using purified Arabidopsis SNARE proteins may represent a good approach to identify the SNARE partners of BET12. Characterization of the interacting domain between BET12 and MEMB12, as well as the use of in vitro budding assays to determine whether BET12 and MEMB12 are co-packaged into vesicles, would certainly shed light on the ER export mechanism of SNARE proteins.

The involvement of the Golgi-localized SNARE MEMB12 in the plant defense against pathogens has been previously reported (Zhang et al., 2011b). Arabidopsis MEMB12 knockout mutants exhibit enhanced resistance to the bacterial pathogen Pst as the absence of MEMB12 promotes the exocytosis of the antimicrobial protein PR1 (Zhang et al., 2011b). Consistent with its role as a negative regulator for PR1 secretion, we found in our study that the MEMB12 overexpression caused intracellular accumulation of PR1. Interestingly, ectopic expression of BET12 affects PR1 trafficking just as MEMB12 does. Although PR1 trafficking was affected, transgenic plants overexpressing BET12 displayed no significant difference in resistance to the Pst infection (Fig. 6), probably due to the fact that the amount of secreted PR1 for defense was not significantly reduced. SNARE proteins are known to play an important role in plant defense (Assaad et al., 2004; Collins et al., 2003; Kwon et al., 2008; Uemura et al., 2012a; Wang et al., 2016). Strikingly, a previous study has revealed the essential role of the plant secretory pathway for the plant immunity, as mutants of the secretory pathway component showed reduced PR1 secretion and were susceptible to bacterial pathogen (Wang et al., 2005). Another study reported that the plasma membrane-localized SNARE SYP132 underwent phosphorylation upon elicitor treatment and the efficient PR1 secretion was SYP132-dependent (Kalde et al., 2007).

Mounting evidence suggests that the absence of certain SNARE proteins or secretory components makes plants more susceptible to pathogens; our finding that the presence of BET12 and MEMB12 play a repressive role in PR1 secretion may therefore seem contradictory. It has been proposed that MEMB12 is involved in retrograde protein trafficking from the Golgi back to the ER, thus PR1 may be recycled back and instead of being secreted (Zhang et al., 2011b). As BET12 shows interaction with MEMB12, it seems reasonable that the overexpression of BET12 would inhibit PR1 secretion by promoting its retrograde transport. Although we cannot rule out this possibility, a possible alternative explanation may be the excess SNARE proteins present due to overexpression. The overabundance of SNARE proteins may result in uncontrolled SNARE partner interaction and thus disrupt the SNARE machinery homeostasis. It has been suggested that SNARE proteins could become non-fusogenic when over-accumulated, and these SNAREs are termed inhibitory SNAREs (i-SNAREs) (Di Sansebastiano, 2013), as evidenced by a study using an in vitro fusion assay (Varlamov et al., 2004). Both the yeast and rat orthologs of BET12, Bet1 and rBET1, show an inhibitory effect on SNARE fusion when in excess (Varlamov et al., 2004). In plants, a previous study proposed that SYP51 and SYP52 behave as the i-SNAREs and affected vacuolar trafficking when they accumulate on the tonoplast (De Benedictis et al., 2013). In that sense, considering our result in PR1 trafficking, it is plausible that overexpressed BET12 could act as an i-SNARE in the early secretory pathway, and therefore prevents the fusion of PR1-containing vesicles and thus inhibit its secretion. Interestingly, overexpression of the Golgi-localized SNAREs SYP31 and MEMB11 strongly inhibited the ER-to-Golgi anterograde transport, as evidenced by the redistribution of Man1–RFP into the ER (Bubeck et al., 2008). It is noteworthy that BET12 overexpression did not affect the distribution of Man1–RFP (Fig. 2A), indicating that the general anterograde transport pathway is not perturbed by BET12. Instead, PR1 trafficking was interfered, suggesting a potential role of BET12 in regulating pathogenesis-related protein secretion and plant immunity. Further studies, combining in vitro SNARE fusion assay with the in vivo monitoring of protein trafficking, will help to characterize i-SNARE activity in plants. It will also open the possibility that, in addition to the well-known fusogenic role of SNARE proteins, the regulation and targeting of non-fusogenic SNAREs to compartments could fine-tune protein trafficking in response to various stress conditions.

It has been recently reported that the single homozygous bet11 or bet12 T-DNA insertional mutants display no obvious vegetative phenotype. Reduced seed set was observed in homozygous and heterozygous bet11/bet12 (and vice versa) double mutants, probably caused by defective pollen tube growth (Bolanos-Villegas et al., 2015). These findings imply that the function of BET11 and BET12 may partially overlap, as functional redundancy has been reported between SNARE family members (Kim and Bassham, 2013; Shirakawa et al., 2010; Uemura et al., 2010). Further studies using either bet12 or bet11/bet12 double mutants in pathogen response will certainly help to elucidate the role of ER-to-Golgi SNAREs in plant immunity.

The cDNA of BET12 was amplified and cloned into the pBI121 backbone containing an UBQ10 promoter (Grefen et al., 2010b), the YFP coding sequence and the nopaline synthase terminator for the generation of YFP–BET12 transgenic plants. pBI221 vectors containing an UBQ10 promoter, the YFP or the HA coding sequence, the BET12 coding sequence and the nopaline synthase terminator were generated for the transient expression in protoplasts. All the truncation and deletion versions and point mutagenesis mutants of BET12 were amplified from YFP–BET12 and cloned into the pBI221 vector. All the primers used in this study are listed in Table S2.

To generate YFP–BET12 transgenic plants, pBI121 constructs containing YFP–BET12 were introduced into Agrobacterium tumefaciens and transformed into wild-type Arabidopsis thaliana (Col-0) by the floral dip method (Clough and Bent, 1998). Arabidopsis seeds were surface sterilized and plated on standard Murashige and Skoog (MS) growth medium (pH 5.7) supplemented with 1% sucrose and 1% agar. Seedlings were grown on vertically oriented plates in growth chambers at 22°C under a long-day (16 h light and 8 h dark) photoperiod. Plants used for the bacteria growth assay were grown on soil in a growth room under a short-day condition with a 10 h light and 14 h dark photoperiod for an extended vegetative growth phase.

Maintenance of Arabidopsis suspension cells and transient expression in protoplasts were performed by electroporation as described previously (Miao and Jiang, 2007). For particle bombardment, 7-day-old Arabidopsis seedlings were transferred and placed horizontally on a MS agar plate. Gold particles were coated with plasmid DNA and bombarded into seedlings as described previously (Wang and Jiang, 2011).

Preparation and fixation of 5-day-old Arabidopsis roots for immunofluorescence labeling was performed as described previously (Sauer et al., 2006). Fixed roots were incubated with anti-EMP12, anti-SYP61 and anti-calreticulin at 4°C overnight, followed by probing with Alexa Fluor 568-conjugated goat anti-rabbit-IgG (Invitrogen) secondary antibody (1:1000 dilution) for confocal observation.

5-day-old Arabidopsis seedlings were treated with BFA at 10 μg/ml for 30 min, followed by FM4-64 uptake for another 30 min before imaging (Lam et al., 2009). All experiments were repeated at least three times with similar results.

CLSM analysis was performed using the Leica SP8 confocal microscope with a sequential line scanning setting using 63× water lens. For each experiment, more than 20 confocal images were collected for quantification and colocalization analysis. The Pearson correlation coefficient was calculated by using the PSC plugin and a line plot was made using ImageJ (Wayne Rasband, NIH, https://imagej.nih.gov/) as described (French et al., 2008).

EM sample preparation, ultrathin sectioning and immunogold labeling using 10-nm gold particles were performed as previously described (Tse et al., 2004). Anti-GFP antibodies were used for labeling of root cells in YFP–BET12 transgenic plants. Transmission electron microscopy was performed using a Hitachi H-7650 transmission electron microscope with a charge-coupled device camera (Hitachi High-Technologies) operating at 80 kV.

Total proteins were extracted from protoplasts expressing YFP–BET12 without the addition of detergent as described (Zhuang et al., 2017). To obtain proteins in soluble and membrane fraction, total proteins were ultracentrifuged at 100,000 g for 30 min, and the membrane pellets were washed and solubilized in an equal volume of extraction buffer with additional 1% Triton X-100. For integral membrane protein determinations, membrane pellets were incubated with 1 M KCl, 0.1 M Na₂CO₃, 1% Triton X-100 and 1% SDS for 30 min on ice, followed by ultracentrifugation at 100,000 g for 30 min. Soluble and membrane fraction were subjected to SDS-PAGE and immunoblot analysis using anti-GFP, anti-VSR and anti-cFPBase antibodies . For protease protection assays, microsomes isolated from protoplasts expressing YFP–BET12 were subjected to trypsin digestion as previously described (Gao et al., 2012), followed by immunoblot analysis using anti-GFP antibodies.

A synthetic peptide of the nine amino acids 98–106 of BET12 was conjugated to CnBr-activated Sepharose as described previously (Contreras et al., 2004). Total proteins were extracted from Arabidopsis suspension cells and were incubated with the conjugated peptide for 4 h at 4°C in a rotator. After incubation, the Sepharose were washed five times with incubation buffer and proteins were eluted by boiling in SDS sample buffer. Proteins were separated by SDS-PAGE and stained by silver staining. Protein bands with significantly higher intensity in the conjugated-peptide lane than in the Sepharose control lane were cut out for in-gel trypsin digestion. Peptides were then extracted from the digested gel and further subjected to liquid chromatography-tandem mass spectrometry analysis as described previously (Gao et al., 2012).

Co-IP assays were performed using proteins extracted from protoplasts expressing YFP–BET12-m and HA–MEMB12. Extracted proteins were incubated with GFP-TRAP magnetic beads following the recommended protocol (ChromoTek, http://www.chromotek.com/). After the washing steps, proteins were eluted and boiled in SDS sample buffer, followed by immunoblot analysis using anti-GFP and anti-HA antibodies. FRET-AB analysis was performed using the Leica SP8 confocal microscope as described previously (Gao et al., 2015). Target proteins were transiently expressed in Arabidopsis protoplasts. Fixation was then performed by incubating the protoplasts with 3% formaldehyde in PBS for 15 min at room temperature. After two rounds of washing with PBS, fixed samples were then subjected to FRET-AB analysis. Defined region of interest was selected for photobleaching using a high-intensity laser (514 nm). The signal intensity of the donor and acceptor proteins before and after photobleaching were measured for calculating FRET efficiency through the built-in SP8 algorithm. For each testing FRET protein pair, 20 individual cells expressing the target proteins were used for the analysis. For the positive control, an amino acid linker peptide, SSSELSGDEVGGTSGSEF, was used to fuse the C-terminus of Cerulean and the N-terminus of YFP to create the Cerulean–linker–YFP fusion, while CNX–Cerulean and YFP–BET12-m were used as a negative control.

Secretion assays were performed using protoplasts expressing PR1–RFP or PR1–RFP co-expressed with YFP–MEMB12 or YFP–BET12. Protoplasts and culture medium were collected separately by centrifugation at 100 g for 5 min. Total proteins were extracted from the protoplasts, while collected medium was concentrated using Amicon^® Ultra-4 Centrifugal Filter Units with a 3 kDa molecular mass cut-off by centrifugation at 3200 g for 30 min. Proteins were extracted and the concentrated medium was subjected to SDS-PAGE followed by immunoblot analysis using anti-RFP and -GFP antibodies. Ponceau S staining was used as a loading reference for calculating the relative abundance of PR1 proteins in each sample.

Bacterial growth assays were performed using 4-week-old wild-type and YFP–BET12 transgenic Arabidopsis plants as described previously (Li et al., 2013). 1×10⁶ cfu/ml of Pst (DC3000) and 5×10⁶ cfu/ml of Pst avrRpt2 were used for infection by infiltration into Arabidopsis leaves with syringes. Eight leaf-discs were collected at 0 dpi and 3 dpi for counting bacterial number. For each growth assay, the results from three biological replicates were used for bacterial quantification.

The anti-GFP antibody (1:1000 dilution) was generated as described previously (Shen et al., 2014). Anti-cFBPase antibody (1:2000 dilution) was purchased from Agrisera (cat. no. AS04043). Anti-HA antibody (1:1000 dilution) was purchased from Abcam (catalog no. ab18181). Anti-RFP antibody (1:1000 dilution) was purchased from chromotek (rat monoclonal 5F8). Anti-VSR, -SYP61, -EMP12 and -calreticulin antibody were used as described previously (Gao et al., 2012; Sanderfoot et al., 2001; Shen et al., 2014; Tse et al., 2004).

Sequence data for the proteins analyzed in this article can be found in the EMBL/GenBank data libraries under the following accession numbers: AT4G14455 (BET12); AT1G10950 (EMP12); AT1G28490 (SYP61); AT2G30290 (VSR2); AF126550 (MAN1); AT5G20990 (CNX); AT4G02080 (SAR1C); AT5G50440 (MEMB12); AT2G14610 (PR1); AT2G28520 (VHA-a1); AT5G05760 (SYP31); AT2G45200 (GOS12); AT2G01770 (VIT1) and OS07G0564600 (SCAMP1).

We gratefully acknowledge Mr Kwok Wai Kwan for technical assistance. We thank Professor Jurgen Denecke (University of Leeds, UK) for providing anti-calreticulin antibodies.