The Golgi entry core compartment functions as a COPII-independent scaffold for ER-to-Golgi transport in plant cells

The Golgi is an essential organelle for membrane trafficking, which functions in protein modification and sorting, and polysaccharide synthesis in plant cells in particular. Its characteristic stacked structure of disk-like cisternae is conserved among almost all eukaryotes, which makes this organelle easily distinguishable from other intracellular structures. The cisternae are polarized between the cis- and trans-sides within a stack, and glycosylation enzymes localize sequentially from cis to trans in the order in which they function (Dunphy and Rothman, 1985; Glick and Luini, 2011). Knowledge about what occurs during the biogenesis of this structure is limited to research on a few protozoans – typically organisms with only one Golgi stack per cell – yet these results have proposed a variety of schemes for Golgi generation (Benchimol et al., 2001; Bevis et al., 2002; He et al., 2004; Pelletier et al., 2002). Molecular mechanisms that govern Golgi formation remain to be unveiled.

Brefeldin A (BFA), a compound that inhibits the activity of guanine nucleotide exchange factors for ARF GTPases (Chardin and McCormick, 1999), has been used as a tool to investigate Golgi assembly in mammalian cells, because its treatment leads to relocalization of Golgi proteins into the ER in a reversible manner (Alcalde et al., 1992; Puri and Linstedt, 2003). Similar disassembly and reassembly of the Golgi is also observed upon BFA treatment and removal in tobacco cells (Langhans et al., 2007; Ritzenthaler et al., 2002; Schoberer et al., 2010). Taking advantage of this, we previously performed multicolor time-lapse observation of Golgi proteins during BFA treatment and removal in tobacco BY-2 cultured cells. We found that previously unknown punctate structures, in which a particular subset of cis-Golgi proteins localize, are formed in the vicinity of the ER exit sites (ERES) upon BFA treatment and that they function as the scaffold for Golgi regeneration after BFA removal, which proceeds in the cis to trans direction (Ito et al., 2012). We propose to call these punctate structures GECCO (the Golgi entry core compartment) hereafter, for the reasons described herein.

Our earlier findings raised several new questions: (1) What is the difference between the cis-Golgi proteins that concentrate to GECCO and those that relocalize to the ER upon BFA treatment? (2) Are the Golgi components that have been relocated to the ER by BFA sent back to the original positions through GECCO after BFA removal? (3) What transport machinery is involved in the formation of GECCO? In this report, we address these questions by using powerful live-imaging techniques.

In our previous work, we found that cis-Golgi proteins SYP31 and RER1B localize to GECCO upon BFA treatment, whereas another cis-Golgi marker ERD2 changes its localization to the ER completely (Ito et al., 2012). In order to determine what is different between these two types of cis-Golgi proteins, we observed their intra-Golgi localization in detail by establishing a BY-2 cell line expressing two cis-Golgi proteins and a trans-Golgi protein (ST) labeled with three different colors simultaneously. As shown in Fig. 1A,B, localization of GFP–SYP31 and ERD2–YFP in comparison to the trans marker ST–mRFP were slightly different. GFP–SYP31 localized to the extreme cis side in the Golgi stacks. GFP–RER1B and mRFP–SYP31 showed mostly overlapping localization that was distinct from ST–YFP (Fig. 1C,D). Although the intra-Golgi localization of ERD2 was relatively broad, its fluorescence peak was subtly different from that of the medial-Golgi marker XYLT (β1,2-xylosyltransferase; Pagny et al., 2003; Fig. S1A,B). This means that the ERD2 signal mainly represents cisternae distinct from the medial-Golgi. In addition, mRFP–SYP31 showed a different intra-Golgi localization to that of ManI (α-1,2-mannosidase I; Fig. S1C,D), which relocated to the ER upon BFA treatment (Fig. S1E). Because ManI functions as the first N-glycosylation enzyme in the Golgi, it was thought to localize at the cis-most side of the stacks. However, an immunolocalization analysis revealed that it localizes at the third to fourth cisternae, not at the first and second cis cisternae (Donohoe et al., 2013). Our observations indicate that SYP31 and RER1B, the cis-Golgi markers that localize to GECCO upon BFA treatment, originally reside in the cis-most region of the Golgi, presumably at the first to second cisternae.

As we reported previously, regeneration of the Golgi stacks after removal of BFA begins by coalescence of GECCO with GFP–SYP31 signals, and ST–mRFP is gradually concentrated next to the regenerating cis-cisternae. However, it was unclear whether ST–mRFP was transported via GECCO (Ito et al., 2012).

To examine the role of GECCO in transport of other Golgi proteins during Golgi regeneration, we observed the Golgi and the ER markers using super-resolution confocal live-imaging microscopy (SCLIM; Kurokawa et al., 2013). SCLIM enables live-cell 3D time-lapse (4D) imaging with extremely high speed and high resolution in multiple colors at the same time, and thus provides us with information that was not previously available. In order to visualize the cis- and trans-Golgi and the ER simultaneously, we used near-infrared fluorescent protein (iRFP) as the third fluorescent protein (Filonov et al., 2011). We constructed a derivative of the fluorescent ER marker SP–iRFP-HDEL, iRFP fused with the signal peptide (SP) from Arabidopsis endo-xyloglucan transferase and the HDEL ER retrieval signal (Takeuchi et al., 2000), and introduced it into a BY-2 cell line already expressing GFP–SYP31 and ST–mRFP. By adding biliverdin (the chromophore of iRFP) to the culture, we could observe the tubular network unique to the ER with Golgi stacks in 3D (Fig. 2). Using this cell line, we observed Golgi regeneration after BFA removal. In order to inhibit the de novo synthesis of marker proteins and to halt the movement of organelles by depolymerizing actin fibers, cycloheximide and latrunculin B (LatB) were added to the culture. ST-mRFP transiently colocalized with GFP–SYP31 at GECCO (Fig. 3A, 2:55–3:55), and then the two fluorescent signals were segregated into cis and trans localizations (Fig. 3A, 5:00–5:35). The 3D images of ST–mRFP and GFP–SYP31 colocalization at GECCO showed a mosaic distribution (Fig. 3B). Pearson's correlation coefficients in 3D between GFP–SYP31 and ST–mRFP exhibited an increase over time and a subsequent decrease to the initial level, whereas that between ST-mRFP and SP-iRFP-HDEL decreased continuously (Fig. 3C, Fig. S2A,B). Such an increase of the correlation between SYP31 and ST was not observed either in the cells without BFA (Fig. 3D) or in cells with continued BFA treatment (Fig. S2C). These data suggest that ST–mRFP that has exited the ER after BFA removal is first transported to GECCO where GFP–SYP31 resides, and then localizes to the regenerated trans-cisternae.

Only a subset of cis-Golgi proteins behave differently from other Golgi proteins and are localized to GECCO upon BFA treatment. We examined whether newly synthesized SYP31 is transported from the ER to GECCO in the presence of BFA. As a control, we transformed BY-2 cells with dexamethasone (DEX)-inducible mRFP-tagged SYP22 (also known as VAM3, a vacuolar membrane SNARE). When DEX and BFA were added to the cell culture at the same time, newly synthesized mRFP–SYP22 did not reach the vacuole but localized to the ER membrane, colocalizing with the ER marker SP–GFP-HDEL (Fig. 4A). This indicates that the ER exit of mRFP–SYP22 was inhibited by BFA treatment. Likewise, when the DEX-inducible mRFP–SYP31 cell line was treated with DEX and BFA simultaneously, mRFP–SYP31 localized to GECCO, similar to the results observed when BFA was added after mRFP–SYP31 induction (Fig. 4B). This indicates that transport of SYP31 from the ER to GECCO was not inhibited by BFA.

Next, we examined whether the COPII machinery is involved in the transport of SYP31 to GECCO. The COPII coat plays a central role in ER-to-Golgi transport, and its assembly is regulated by SAR1 GTPase (Brandizzi and Barlowe, 2013; Sato and Nakano, 2007). Expression of the GTP-fixed mutant form of SAR1 GTPase (SAR1 H74L) is known to have a dominant-negative effect on ER–Golgi transport (Osterrieder et al., 2009; Takeuchi et al., 1998, 2000). We introduced DEX-inducible NtSAR1 H74L into the cell line stably expressing GFP–SYP31 and ST–mRFP. As shown in Fig. 5A, the induction of NtSAR1 H74L led to the localization of ST-mRFP to the ER and of GFP–SYP31 to GECCO, similar to results with BFA treatment. The fluorescence profile around GECCO was markedly different from that in the control Golgi. ST–mRFP on the ER did not show apparent accumulation near GECCO (compare Fig. S3A,B), indicating that ST–mRFP is almost evenly dispersed on the ER. In order to examine ER-to-GECCO transport, we established a cell line in which GFP–SYP31 and ST–mRFP can be concurrently induced by estradiol treatment (Fig. 5B), and further transformed it with DEX-inducible NtSAR1 H74L. After 24 h of DEX treatment, the same condition that caused relocalization of stably expressed ST–mRFP, we added estradiol to induce the Golgi markers. The newly expressed ST–mRFP localized to the ER, and GFP–SYP31 showed GECCO localization (Fig. 5C). The fluorescent profile showed no obvious accumulation of ST–mRFP near GECCO (Fig. 3C). These data indicate that the localization of SYP31 to GECCO does not depend on the COPII machinery.

In this study, we found that the proteins that localize to what we propose to call GECCO upon treatment with BFA, are normally localized to the cis-most region of the Golgi stack (Fig. 6A,B). These observations suggest that the cis-most cisternae have a different property from the rest of the Golgi cisternae in plant cells.

In vertebrate cells, in which Golgi stacks are concentrated near the nucleus to form a large ribbon-like structure, the Golgi is far from the ERES (Wei and Seemann, 2010). In order to carry out this long-distance transport, the ER-Golgi intermediate compartment (ERGIC) functions as a cargo-sorting station between the ER and the Golgi (Appenzeller-Herzog and Hauri, 2006). By contrast, in plant cells, the Golgi stacks and the ERES are always closely associated and therefore ERGIC was believed to be nonexistent in plants. In our previous report, however, we noticed the similarity between the punctate cis-Golgi compartment formed upon BFA treatment (which we now call GECCO) in plant cells and the ERGIC in vertebrate cells, because they both act as the scaffold for Golgi formation/regeneration (Ito et al., 2012). We speculated that this special cis-Golgi compartment quickly matures or merges into cis-cisternae under normal conditions and is exaggerated and becomes visible when their maturation is arrested by BFA.

Now, taking the new data into consideration, we would like to update our model. An ERGIC marker ERGIC-53 is known to localize to punctate structures near the ERES upon BFA treatment in mammalian cells, which are called ‘Golgi remnants’ (Lippincott-Schwartz et al., 1990). GECCO, the punctate structure of cis-Golgi markers formed adjacent to the ERES upon BFA treatment in plant cells, is more like these Golgi remnants of mammalian cells. SYP31 and RER1B, which localize to GECCO in the presence of BFA, normally localize to the cis-most region of the Golgi. We propose that the cis-most cisternae in plant cells originally have a property corresponding to the ERGIC, and harbor a functional domain to receive cargo from the ER (GECCO). An immuno-electron microscopy study has also suggested that the cis-most cisternae in plant cells are biosynthetically inactive and function as sorting compartments, similar to ERGIC (Donohoe et al., 2013). Plants and vertebrates share the fundamental machinery at the ER-Golgi interface, albeit with different spatial arrangements.

Because our previous study provided evidence for the role of GECCO as the scaffold for Golgi reassembly, we now aimed to observe the transport processes at this scaffold in 3D in detail. The high-speed and high-sensitivity SCLIM, a super-resolution live-imaging microscopy we have developed (Kurokawa et al., 2013), now enables us to make 4D observations with triple colors at low laser powers, which avoids the photodamage that can halt Golgi regeneration.

In the 4D data, we see temporary colocalization of ST–mRFP with GFP–SYP31 at GECCO during regeneration. Correlation coefficients between these markers increase during this step, and then decrease as typical side-by-side localization of cis- and trans-Golgi appears. The temporal increase of colocalization is small in extent but is significant and reproducible (see Figs 2 and 3). These results indicate that ST–mRFP is initially transported to GECCO, rather than directly localizing to the reassembling trans-cisternae (Fig. 6C,D).

During the colocalization, GFP–SYP31 and ST–mRFP are not mixed evenly and show mosaic patterns. This is quite similar to the distribution patterns of earlier and later Golgi markers coexisting within one maturing cisterna in Saccharomyces cerevisiae (Ishii et al., 2016; Matsuura-Tokita et al., 2006). Although the mechanism of this domain formation is yet to be investigated, it is reasonable that such segregation of different Golgi proteins contributes to their efficient sorting and transport. The similar segregation of cis- and trans-Golgi markers at GECCO after removal of BFA presumably indicates that the Golgi stacks regenerate by using the mechanism of cisternal maturation.

COPII is the central mechanism of the ER-to-Golgi anterograde transport, and its disruption by dominant-mutant forms of SAR1 is well known to arrest ER exit of many kinds of cargo in yeast (Nakano et al., 1994), mammals (Kuge et al., 1994) and plants (reviewed in Ito et al., 2014). Indeed, in the present study, we have shown that the ER exit of ST–mRFP is blocked by the SAR1 H74L mutant. In mammalian cells, ERGIC-53 is trapped in the ER by dominant-negative SAR1 (Hauri et al., 2000; Shima et al., 1998). However, in tobacco leaf epidermal cells, cis-Golgi matrix proteins are reported to localize not only to the ER but also to small punctate structures upon expression of NtSAR1 H74L, depending on the cells and transformation methods used (Osterrieder et al., 2009). The authors suggested that this variation of localization is due to the difference in expression timing between the fluorescent Golgi proteins and NtSAR1 H74L; the Golgi proteins that had already been exported from the ER before blockage by NtSAR1 H74L were left at the punctate structures after ER-exit arrest. In our present study, we do not observe any cells with GFP–SYP31 solely localizing to the ER upon induction of NtSAR1 H74L. SYP31 is still transported to GECCO when NtSAR1 H74L is induced sufficiently earlier than Golgi markers by time-shift dual induction. Therefore, we propose that a ‘core’ compartment for Golgi entry (GECCO) exists with a particular set of Golgi proteins (components of the cis-most cisternae) in plant cells, the formation of which is independent of the COPII mechanism (Fig. 6B). Although COPII-independent events in the ER-to-Golgi traffic are not well understood and may be provocative, our findings provide a good opportunity to consider whether this is a plant-specific phenomenon or could be applicable to other organisms such as mammals and yeast. Obviously, more extensive analysis on proteins trafficking across ER–GECCO–ERGIC–Golgi will be necessary to bring us further insights.

As described above, we defined the compartment where a particular set of cis-Golgi proteins accumulate upon BFA treatment as the Golgi entry core compartment (GECCO). We would like to further extend this definition. GECCO can exist under normal conditions, maybe quickly maturing or merging into the cis-most cisternae, and receive COPII vesicles at the ER–Golgi interface. In other words, its important role is to function as the entry compartment of the Golgi generation, existing as a functional domain in the cis-most cisternae.

In S. cerevisiae, observations by SCLIM have revealed that ER-to-Golgi transport is mediated by the hug-and-kiss action of the cis-Golgi; it approaches and contacts the ERES to receive cargo (Kurokawa et al., 2014; Robinson et al., 2015). If this mechanism also operates during formation of new cis cisternae, some special compartment needs to pre-exist in front of the ER to receive components to build cis-Golgi. In the classical cisternal maturation model, the cis-most cisterna is assumed to form by coalescence of ER-derived COPII vesicles and perhaps COPI vesicles (Glick and Luini, 2011; Nakano and Luini, 2010); however, this is not consistent with the pre-existence of the receiver compartment that fulfills the hug-and-kiss action. We suggest that a special cis-Golgi compartment in yeast plays a role as GECCO and contributes as the receiver for the hug-and-kiss transport. Whether it is formed independently of COPII is a very interesting question that should be addressed.

COPII is the only mechanism known to date at the molecular level to be responsible for the ER-to-Golgi anterograde trafficking. Although there are several reports arguing for COPII-independent cargo transport from the ER in yeast, Drosophila and mammalian cells, such systems appear to bypass the Golgi after the exit from the ER (Grieve and Rabouille, 2011; Rabouille, 2017). A study on mammalian cells reported that transport of procollagen-I was inhibited by silencing of Sar1 but that of VSV-G and albumin was not (Cutrona et al., 2013). Under these conditions, the typical spatial organization of the ER–Golgi boundary was impaired, but mini Golgi stacks were observed in the vicinity of the ER. The authors suggested that COPI played a role in bypassing COPII (Cutrona et al., 2013). In our present study, as SYP31 exits the ER and localizes to GECCO in the presence of BFA, the involvement of the COPI machinery is unlikely in this COPII-independent pathway. Elucidation of the molecular mechanisms of this novel traffic route and the formation of GECCO would impact our understanding about Golgi generation.

To generate ERD2-YFP, ST-YFP, and ManI-YFP constructs, DNA fragments coding for ERD2, ST and ManI were amplified by PCR and cloned into pENTR/D-TOPO (Thermo Fisher Scientific/Invitrogen, Waltham, MA), and recombined into pK7YWG2 (Karimi et al., 2005) by LR Clonase II (Thermo Fisher Scientific/Invitrogen). ERD2-mRFP construct was generated by recombining ERD2 in pENTR/D-TOPO into pH7RWG2 (Karimi et al., 2005). For SP-iRFP-HDEL, the GFP region of SP-GFP-HDEL (Takeuchi et al., 2000) was replaced by the DNA fragment coding for iRFP by In-Fusion HD Cloning Kit (Takara Bio/Clontech, Shiga, Japan). XYLT-GFP is a kind gift from Keiko Shoda (RIKEN, Japan).

For the DEX-inducible constructs, we modified pTA7002 (Aoyama and Chua, 1997). The T-DNA region (from RB to LB) of pKGW (Karimi et al., 2002) was replaced with the T-DNA region of pTA7002, and the sequence encoding the hpt gene in the T-DNA region of this vector was replaced with the nptII gene by In-Fusion. Next, the DNA fragment of the attR1–attR2 region of pHGW (Karimi et al., 2002) was inserted into the XhoI site. The resulting plasmid was designated pTASKGW (made from pTA7002, Spe^r in bacteria and Km^r in plants, with a Gateway cassette). To generate DEX-inducible fluorescent markers, the DNA fragments coding for mRFP-SYP22 and mRFP-SYP31 were amplified by PCR and cloned into pENTR/D-TOPO, followed by recombination into pTASKGW by LR Clonase II. For the construction of DEX-inducible NtSAR1 H74L, the DNA fragment encoding NtSAR1 (GenBank accession number: D87821) was amplified from BY-2 cDNA by PCR, and cloned into pENTR/D-TOPO. To introduce a mutation, the whole vector was amplified by PCR using a mutagenesis primer set (5′-GGAGGTCTTCAGATCGCTCGCCGTGTC-3′ and 5′-GATCTGAAGACCTCCTAAATCAAACGCTTTGAA-3′) by PrimeSTAR Max DNA Polymerase (Takara Bio), and directly transformed into E. coli to obtain a self-circularized plasmid with NtSAR1 H74L. This was subcloned into pTASKGW by LR Clonase II.

To generate estradiol-inducible GFP-SYP31/ST-mRFP, DNA fragments coding for GFP-SYP31 and ST-mRFP were first cloned into pMDC7 (Curtis and Grossniklaus, 2003). The DNA region from O_lexA-46 to T_3A of pMDC7 harboring GFP-SYP31 was amplified by PCR, and inserted into the KpnI site of pMDC7 harboring ST-mRFP.

Maintenance of tobacco (Nicotiana tabacum) bright yellow-2 (BY-2) culture is described in Nagata et al. (1992). Transformation procedures are described in Ito et al. (2012).

BFA (50 µM) treatment and removal, and LatB (2 µM) and cycloheximide (100 µM) treatments are described in Ito et al. (2012). For gene induction, 10 mM stock solution of dexamethasone (Wako, Tokyo, Japan) diluted in DMSO was added to suspension cultures at 10 µM in the final concentration. Similarly, 20 mM stock solution of β-estradiol (Sigma-Aldrich, St Louis, MO) diluted in DMSO was used at 20 µM final concentration. For fluorescence observation of iRFP, biliverdin hydrochloride (Frontier Scientific, Logan, UT) diluted in DMSO at 25 mM was used at 25 µM in the final concentration. The timings when these drugs were added are indicated in figure legends.

2D triple-colored observations and double-colored observations for Fig. 1 were made under a confocal laser-scanning microscope (model LSM780; Zeiss, Jena, Germany). 2D single- and double-colored observations except for Fig. 1 were also done under a BX52 microscope (Olympus, Tokyo, Japan) equipped with a confocal scanner unit (model CSU10; Yokogawa Electric, Tokyo, Japan) and a cooled digital CCD camera (model ORCA-ER; Hamamatsu Photonics, Shizuoka, Japan). 3D imaging was performed by the SCLIM system we developed (Kurokawa et al., 2013), consisting of the Olympus model IX-71 fluorescence microscope with a custom-made high-speed confocal scanner (Yokogawa Electric), image intensifiers (Hamamatsu Photonics) with a custom-made cooling system, and EM-CCD cameras (Hamamatsu Photonics). The objective lens was oscillated vertically by a custom-made piezo actuator system (Yokogawa Electric). The cells were placed on 35 mm glass-bottom dishes with poly-L-lysine coating (Matsunami, Osaka, Japan), and incubated on a thermos-controlled stage (Tokai Hit, Shizuoka, Japan) maintained at 28°C. Data were subjected to deconvolution analysis with Volocity (Perkin Elmer, Waltham, MA) using the theoretical point-spread function for the spinning-disk confocal system. Images were processed and analyzed with ImageJ 1.49i (National Institutes of Health, Bethesda, MD), Photoshop CS6 (Adobe System, San Jose, CA), and Volocity.

We thank T. Nakagawa (Shimane University, Japan), N.-H. Chua (Rockefeller University, USA) and K. Shoda (RIKEN, Japan) for sharing materials.