A missense mutation in the vacuolar protein GOLD36 causes organizational defects in the ER and aberrant protein trafficking in the plant secretory pathway


(fax +1 517 353 916; e-mail fb@msu.edu).


A central question in cell biology is how the identity of organelles is established and maintained. Here, we report on GOLD36, an EMS mutant identified through a screen for partial displacement of the Golgi marker, ST-GFP, to other organelles. GOLD36 showed partial distribution of ST-GFP into a modified endoplasmic reticulum (ER) network, which formed bulges and large skein-like structures entangling Golgi stacks. GOLD36 showed defects in ER protein export as evidenced by our observations that, besides the partial retention of Golgi markers in the ER, the trafficking of a soluble bulk-flow marker to the cell surface was also compromised. Using a combination of classical mapping and next-generation DNA sequencing approaches, we linked the mutant phenotype to a missense mutation of a proline residue in position 80 to a leucine residue in a small endomembrane protein encoded by the gold36 locus (At1g54030). Subcellular localization analyses indicated that GOLD36 is a vacuolar protein and that its mutated form is retained in the ER. Interestingly also, a gold36 knock-out mutant mirrored the GOLD36 subcellular phenotype. These data indicate that GOLD36 is a protein destined to post-ER compartments and suggest that its export from the ER is a requirement to ensure steady-state maintenance of the organelle’s organization and functional activity in relation to other secretory compartments. We speculate that GOLD36 may be a factor that is necessary for ER integrity because of its ability to limit deleterious effects of other secretory proteins on the ER.


A hallmark of eukaryotic cells is the presence of diverse organelles that perform specialized functions to ensure the cell’s survival. Many of the organelles are interconnected in a functional pathway by extensive exchange of membranes and lumenal contents, all finely tuned for steady-state maintenance of the functional and morphological identity of each organelle within the pathway. The secretory pathway offers a prime example of such complexity. The gateway of the secretory pathway is the endoplasmic reticulum (ER), which provides synthesis, modification, and quality control of secretory proteins (Vitale and Denecke, 1999). Properly folded proteins move to the Golgi apparatus for further modification prior to reaching the trans-Golgi network for sorting to the extracellular space or storage and lytic organelles (Foresti and Denecke, 2008).

The organelles involved in the pathway accomplish their functional tasks and communicate with other secretory organelles while maintaining a morphological and functional identity. How this is achieved is a fascinating question that is largely unanswered, especially in the plant system. Interestingly, secretory organelles can remodel their structure over time and yet maintain their functional and morphological independence from other organelles in the same pathway. For example, the ER is a major intracellular membrane system that consists of a single interconnected dynamic network of membranes (Brough et al., 2005; Borgese et al., 2006). In plant cells, the ER forms a distinct network at the cell periphery (the cortical ER) that extends into the cell (to the cytoplasmic ER) across trans-vacuolar strands (Knebel et al., 1990; Sparkes et al., 2009b). The network undergoes rearrangements due to extensive motility of the ER membranes and continuous formation/absorption of ER tubules and cisternae (Sparkes et al., 2009a). The ER maintains its morphological and functional integrity despite being attached to the Golgi apparatus in highly vacuolated cells (Sparkes et al., 2009c). The Golgi apparatus, on the other hand, exchanges cargo continuously with the ER and is continuously renovated through the balanced action of anterograde and retrograde pathways at the ER/Golgi interface (Brandizzi et al., 2002b; Sparkes et al., 2009c).

Interference with the relative content of membrane proteins in the ER or activities of cytosolic machineries involved in protein transport at the ER/Golgi interface results in a drastic reshaping of the ER network. For example, conditions of over-expression of certain membrane proteins or interference with the activity of COPII and COPI machineries for ER/Golgi protein trafficking leads to a shift to a more cisternal form of ER over tubular ER (daSilva et al., 2004; Runions et al., 2006; Stefano et al., 2006). It has also been shown that ER tubules can form large, convoluted, globular formations at the cell periphery in a mutant bearing a non-functional COPII coat component involved in selective export of membrane proteins, specifically AtSec24A (Faso et al., 2009; Nakano et al., 2009). This mutant was mapped in the cargo binding site of AtSec24A and the mutant phenotype was interpreted as if defects in COPII may result in the inability of the ER to export specific membrane proteins necessary for the maintenance of the dynamic integrity of the ER network (Faso et al., 2009; Nakano et al., 2009). These observations further support the possibility that ER integrity depends on protein export functions of this organelle.

To learn more about the elements involved in the early secretory pathway, we have used a confocal microscopy-based approach to find ethyl methanesulfonate (EMS) mutants with obvious alteration of the distribution of a fluorescent marker destined to the Golgi apparatus (Boulaflous et al., 2008). Because the Golgi is a central organelle for protein traffic in the secretory pathway, defects in Golgi protein distribution in the mutants allow the identification of mutants with defects in Golgi morphology and/or defects in protein transport to and from the Golgi (Boulaflous et al., 2008). Here, we describe the identification and characterization of a recessive EMS mutant showing partial distribution of the Golgi marker ST-GFP (Boevink et al., 1998) to the ER. We showed that the mutant, named GOLD36 (for Golgi defects 36), had defects in both general ER protein export and ER organization. The mutation responsible for the phenotype was pinpointed through a combination of classical mapping procedures and deep sequencing of the mutant genome. We established that the phenotype was linked to a missense mutation in a protein encoded by the At1g54030 locus [mvp1, gtt22 (Nagano et al., 2008; Agee et al., 2010)]. We established that a functional fluorescent fusion of GOLD36 was targeted to the vacuole, whereas its mutant form was retained in the ER. These results indicate that lack of GOLD36 export from the ER, rather than retention of its mutated form in this organelle, is at the basis of defects in functional and morphological integrity of the ER. These data demonstrate that ER morphological and functional integrity in plants depends not only on ER membrane protein distribution and cytosolic determinants that bind the ER surface, but also on distally targeted secretory proteins.


Identification of a mutant with altered distribution of the Golgi marker, ST-GFP

In a microscope-based screen of cotyledons of 1-week-old Arabidopsis seedlings for altered subcellular distribution of ST-GFP (Figure 1a1), a well-established trans-Golgi membrane marker (Boevink et al., 1998), we isolated GOLD36, a mutant with an ST-GFP signal distributed to the Golgi and an underlying network (Figure 1a2). The network defined unusual structures in epidermal cells (Figure 1a2,b); compare Movie S1a and Movie S1b). Generally, in the mid to lower sections of each cell we found one large globular structure (≤10 μm) that contained Golgi stacks (Figure 1a2,b) and that was distinguishable from the nucleus (Figure S1), and also smaller circular structures at the cell’s periphery (Figure 1a2,b); Movie S1b). Both types of structures were highly motile (Figure 1b; Movie S2).

Figure 1.

GOLD36 shows clear defects in the distribution of the Golgi membrane marker, ST-GFP.
(a1) Confocal optical sections (numbers at the left corner indicate depth of sections in microns) of cotyledons of non-mutagenized ST-GFP (n.m. ST-GFP, control) from the cell cortex (0.0 μm) to mid-section (12.0 μm) showing that the fluorescence of the marker is distributed at Golgi stacks (arrows). The full optical section sequence at a 1-μm step interval is reported in Movie S1a. (a2) In contrast to the control, the ST-GFP marker in the GOLD36 mutant is partially distributed to Golgi stacks (0.0 μm section, arrow) and a network (0.0 μm section, arrowheads). ST-GFP is distributed also to circular structures (6.0 μm frame, empty arrows) and large globular structures (9.0 and 12.0 μm frames, stars). The full optical section sequence at a 1-μm step interval is reported in Movie S1b. (b) Images from a time-lapse sequence (from Movie S2) showing that the circular and globular structures labeled are motile in GOLD36 cells. Insets: magnification of areas of interest. Bars = 5 μm (a), 20 μm (b), 2 μm (a2 inset), 10 μm (b inset).

As evidenced by treatment with the general plant endomembrane dye DiOC6 (Zheng et al., 2004; Faso et al., 2009), the phenotype was present also in segregating plants of the F2 mapping population from crosses of the GOLD36 mutant with Arabidopsis thaliana Landsberg erecta (Ler) ecotype that did not express ST-GFP, thereby excluding the possibility that the phenotype was due to over-expression of the transgene (Figure S2).

Analyses of T1 GOLD36 cotyledon epidermal cells stably expressing G-yk, a cis-Golgi marker (Nelson et al., 2007), showed that the fluorescence pattern of ST-GFP overlapped that of G-yk (Figure 2), indicating an abnormal distribution of another Golgi marker of earlier cisternae than the trans-Golgi in GOLD36 and excluding the possibility that the partial redistribution of a Golgi marker was specific to ST-GFP and trans-Golgi proteins.

Figure 2.

 The GOLD36 structures are not specific to ST-GFP.
Two confocal optical slices [(a), cortical region and (b), 2 μm deeper than (a)] of cotyledon epidermal cells stably expressing the cis-Golgi marker G-yk either in the Col-0 (a1, b1 control) or GOLD36 background (a2, b2). Similarly to the control, in GOLD36 G-yk is clearly distributed to Golgi stacks (a, b arrowheads). In GOLD36, G-yk is also partially redistributed to a cortical network (a2, arrow) and to the same structures labeled by ST-GFP, such as a large globular structure (b2, empty arrows) and smaller circular structures (b2, arrows). Insets in b2: magnification of areas of interest. Bars = 10 μm (main image); 5 μm (inset).

Because the membranous network highlighted by the Golgi markers in the GOLD36 mutant background resembled that of the ER (Figures 1 and 2), we analyzed T1 GOLD36 cotyledons expressing the ER lumenal marker, ER-yk (Nelson et al., 2007). Confocal microscope analyses confirmed that the network was indeed ER (Figure 3) and, thus, that in the GOLD36 mutant the Golgi markers are partially redistributed to this organelle.

Figure 3.

 The network highlighted by the Golgi markers in GOLD36 corresponds to ER.
Two confocal optical slices [a, cortical region and b, 20 μm deeper than (a)] of cotyledon epidermal cells stably expressing ER-yk, which labels the ER, either in the Col-0 [control, (a1, b1), arrows] or GOLD36 (a2, arrows) background show that the network highlighted by ST-GFP corresponds to ER. The ER is part of the GOLD36 aberrant structures (b2, arrows and empty arrows). In (a1), arrowheads point to ER fusiform bodies which are also labeled by ER-yk, as also reported for other ER soluble and secreted markers (Hawes et al., 2001; Hayashi et al., 2001; Estevez et al., 2006; Faso et al., 2009). Bar = 10 μm.

The imaging of the smaller circular GOLD36 structures indicated that they were part of the ER network (Figures 1–3; Movies S1 and S2); we aimed to test whether the larger globular structures were in continuity with the ER or formed an isolated compartment. To this end, we photobleached the ER-yk fluorescence in the large globular structure and followed its recovery over time. As shown in Figure 4, ER-yk photobleaching was followed by fluorescence recovery albeit to a much slower recovery rate than previously described for lumenal ER GFP (Tolley et al., 2008), indicating not only that the large globular structures did not form a separate compartment from the general ER, but also suggesting that the defects in ER organization may limit diffusion of proteins in this organelle. The obvious defects in ER morphology raised the question whether the ER export activities of this organelle would also be compromised.

Figure 4.

 The large globular structures are not isolated bodies. Images from a FRAP experiment on the ER-yk fluorescence in a large globular structure (arrows) in a GOLD36 cotyledon epidermal cell stably expressing the ER marker show fluorescence recovery upon selective bleaching. Cells were treated with latrunculin B (25 μm) (Brandizzi et al., 2002b) to reduce the structure’s movement. Only the signal from the YFP channel is shown for clarity. Time of acquisition of single frames is indicated in seconds (s) at the top left corner. Bar = 5 μm.

The GOLD36 mutant has defects in ER protein export

The evidence that two Golgi markers were partially distributed to the GOLD36 ER (Figures 1 and 2) suggested defects in ER protein export. To test this possibility, we aimed to establish the distribution of a bulk-flow reporter such as a soluble marker destined to the apoplast [sporamin signal sequence-mRFP fusion, secRFP (Faso et al., 2009)]. In Col-0, the marker reached its destination (Figure 5a1,b1), as expected (Faso et al., 2009). However, analyses of cotyledon epidermal cells of T1 GOLD36 seedlings expressing secRFP showed partial retention of the marker in the ER network (Figure 5a2,b2), in clear contrast with the control. Interestingly, we also found that in GOLD36, secRFP was retained within the large globular structure and showed evident signal within the smaller circular structures at the cell’s periphery (Figure 5a2,b2). Combined with the evidence of a redistribution of Golgi markers in the ER (Figures 1 and 2), partial retention of secRFP in the ER supports our hypothesis that the GOLD36 mutant has defects in ER protein export.

Figure 5.

 In addition to a partial distribution of Golgi markers in the ER, GOLD36 also shows defects in ER export of a soluble secretory marker destined to the apoplast.
Two sequential confocal optical slices of either control [Col-0 transformed with secRFP (a1, b1), or GOLD36 cotyledon epidermal cells transformed with the soluble apoplastic marker, secRFP (a2, b2)]. The marker is clearly visible in the apoplast in control and GOLD36 cells (arrows); however, in GOLD36 cells, the marker was also partially retained in the ER network labeled by ST-GFP (empty arrows) where it was distributed in bright circular structures (arrowheads), as well as in the globular structures (stars). Insets in (a2): magnification of areas of interest. Bars = 5 μm (b), 2 μm (inset).

The GOLD36 aggregations contain large, striated structures in the ER lumen

We next analyzed GOLD36 at an ultrastructural level. Transmission electron microscopy revealed that the large structures observed in the cotyledon epidermal cells of GOLD36 seedlings were in fact parallel aggregations of large inclusions inside the lumen of the ER (Figure 6a,b). Interestingly, these structures appeared to be semi-crystalline in nature, with longitudinal striations (Figure 6c) and were considerably thinner (50–300 nm in diameter) than the typical ER-located fusiform bodies (300–1200 nm in diameter; Figure 6e) reported previously in Arabidopsis (Hawes et al., 2001; Matsushima et al., 2003), and on occasion they showed a branching pattern (Figure 6c). The typical fusiform bodies with less discernable internal structure were also found in the ER outside the aberrant globular structures (Figure 6d). There were no obvious differences in the organization of other organelles in these cells.

Figure 6.

 Ultrastructure of aberrant structures in GOLD36 cotyledon epidermal cells.
(a) Image of a globular structure showing numerous striated inclusions. Bar = 2 μm. (b) A globular structure lying orthogonal to the cell showing same inclusions. Bar = 1 μm. (c) Higher magnification image of (a) showing detail of the elongate striated structures, some of which branch (arrows). Bar = 500 nm. G = Golgi body. Inset shows detail of longitudinal striated patterning of the inclusions. Bar = 50 nm. (d) Typical ER fusiform body in the cortical ER of a leaf epidermal cell. Notice lack of striations. Bar = 500 nm. (e) ER fusiform bodies in a wild-type cotyledon epidermal cell showing the same structure as that in (d). Bar = 500 nm.

The GOLD36 phenotype is linked to a mutation in the At1g54030 locus

We believe that GOLD36 represents a mutant with clear ER morphological and functional defects. Therefore, we next sought to identify the mutation responsible for the GOLD36 phenotype. To generate a mapping population, we crossed the GOLD36 mutant with Ler. The genomic DNA of 57 F2 plants exhibiting the recessive phenotype was analyzed and a candidate region with the mutation was rough mapped between F11F12 and F14J16 BAC clones on chromosome 1 at a 2.2-Mbp intervals. To identify the point mutation, we sequenced the genomic DNA of the GOLD36 mutant by Illumina Genome Analyzer II (Solexa). Within the candidate region, 147 homozygous single nucleotide polymorphisms (SNPs) were identified, of which 18 were found within transcription units. Seven of these SNPs were found within coding regions, but only four were predicted to be non-silent mutations (Table S1). Among these, only one SNP resulted in a typical G/C-to-A/T EMS transition (Maple and Moller, 2007), which caused a CCT→CTT codon change in the At1g54030 locus (herein named gold36) and resulted in the missense mutation of a proline residue in position 80 into a leucine residue (P80L) at transcriptional level (Figure 7a). To test whether such mutation was indeed responsible for the observed phenotype, we analyzed T1 GOLD36 seedlings stably expressing the cDNA of either gold36 or gold36P80L under the control of a CaMV 35S promoter using a confocal microscope. We found that the GOLD36 phenotype was complemented by the wild-type cDNA but not by the cDNA bearing the mutation (Figure 7b,c), indicating that the P80L missense mutation was indeed responsible for the phenotype.

Figure 7.

 The mutation responsible for the GOLD36 phenotype resides in the At1g54030 locus.
(a) Diagram of the At1g54030 locus showing exons (filled boxes), the position of CCT→CTT EMS responsible for the P80L mutation (gold36P80L) and the insertion site of the T-DNA in the gold36–1 line. (b) RT-PCR to amplify wild type and mutated gold36 cDNA (1251 bp; top images) and UBQ10 (PCR control, 379 bp; bottom images) in non-mutagenized ST-GFP (control) and GOLD36 backgrounds showing higher levels of transcripts in the GOLD36 lines compared to the control, as expected for transformed lines. PCR cycles are indicated below the RT-PCR panels. (c) Confocal images of cotyledon epidermal cells of either control plants or GOLD36 transformed with wild-type and mutant gold36, showing that the GOLD36 phenotype is complemented by gold36 but not by its mutant sequence. Arrow points to an aberrant GOLD36 structure. Bar = 20 μm.

A gold36 knock-out mutant phenocopies the GOLD36 phenotype

Because we were working with a missense mutant, we aimed to compare the phenotype with that of a knock-down/out mutant. This experiment was important to distinguish whether the ER integrity defects of the GOLD36 mutant were linked to the presence of the GOLD36 mutant protein or to gold36 reduced/absent transcript levels. To do so, we isolated a T-DNA Col-0 insertion line (SALK_030621) in the first exon from the ABRC stock center (gold36–1;Figure 7a; Figure S3a). RT-PCR of homozygous lines for the T-DNA insertion did not show presence of transcript compared to controls (Col-0, non-mutagenized ST-GFP, and GOLD36 backgrounds) (Figure 8a), indicating that gold36–1 is most likely a knock-out.

Figure 8.

 A gold36 knock-out mutant phenocopies the GOLD36 phenotype.
(a) RT-PCR analysis for the gold36 transcript in non-mutagenized ST-GFP, Col-0, GOLD36, and gold36–1 backgrounds showing non-appreciable levels of gold36 in the gold36–1 background compared to the others, and suggesting that the line is most likely a knock-out. Top panel: gold36; bottom panel: UBQ10. PCR cycles are indicated below the RT-PCR panels. (b) Three sequential confocal optical slices of gold36–1 cotyledon epidermal cells stably expressing ST-GFP. Note the partial distribution of ST-GFP in the ER (panel 0.0 μm, arrow) besides Golgi stacks (panel 0.0 μm, arrowheads) and the presence of aberrant circular (panel 1 μm, arrows) and large globular (panel 8 μm, empty arrow) structures in the ER. Bar = 20 μm.

To test whether the GOLD36 phenotype was linked to an aberrant protein product (i.e. GOLD36P80L) or abnormal gold36 transcript levels, we aimed to compare the gold36–1 subcellular phenotype with that of GOLD36. To this end, we transformed gold36–1 with ST-GFP. We found that the distribution of ST-GFP in transformed gold36–1 T1 cotyledons clearly phenocopied that of GOLD36, in that ST-GFP was partially redistributed to the ER, as well as to globular and circular structures (compare Figures 1 and 8b). In addition, complemented lines (gold36–1 transformed with 35S::gold36) did not have such structures (Figure S3b). Together, these data confirm our forward genetics analyses showing that the GOLD36 phenotype was linked to the gold36 locus. Importantly also, the evidence that a knock-out mutant phenocopied GOLD36 suggested not only that GOLD36P80L is a non-functional protein, but also indicated that the ER integrity defects of the GOLD36 mutant are not linked to the presence of the aberrant protein gold36 product in the ER but rather to the absence of intact gold36 transcript.

GOLD36 is targeted to the vacuole, but GOLD36P80L is retained in the ER

Analyses of the primary sequence of GOLD36 indicate that the protein contains 417 amino acid residues with a predicted molecular mass of approximately 46 kD. GOLD36 is annotated as MVP1 (Agee et al., 2010) and as a member of the superfamily of plant GDSL-like lipases; however, unlike all the other members, it does not contain the flexible active site serine in the canonical GDSL motif located near the N-terminus that is essential for lipase activity in this family (Brick et al., 1995). In accordance with this finding, absence of lipase activity of GOLD36/MVP1 has been confirmed recently (Agee et al., 2010). SignalP 3.0 prediction server (Emanuelsson et al., 2007) indicates the presence of an N-terminal signal sequence for insertion in the ER. The protein sequence does not have obvious retention signals, suggesting that it may be exported from the ER to distal compartments such as the vacuole and the apoplast. To investigate this possibility, we aimed to examine GOLD36 localization using a stable transformation system. We first transformed GOLD36 plants with a GOLD36 fusion to a monomeric red fluorescent protein (mRFP; 35S::gold36-mRFP). mRFP is more resistant to low pH than GFP (Shaner et al., 2005); therefore, it is better suited as a marker for acidic compartments such as the apoplast or the vacuole than GFP. Confocal analyses showed that the GOLD36-mRFP was clearly localized in the vacuole (Figure 9a). GOLD36-mRFP complemented the GOLD36 phenotype (Figure 9a), indicating that the protein is a functional fusion; therefore, its distribution must at least partially overlap with that of the endogenous GOLD36. Because the mRFP signal was confined to the cell body rather than in the apoplastic space in plasmolyzed GOLD36 cotyledonal epidermal cells expressing GOLD36-mRFP, we concluded that GOLD36-mRFP is not secreted to the apoplast (Figure S4). Further evidence that GOLD36 is directed to the vacuole is provided in Figure S5.

Figure 9.

 GOLD36 is a vacuolar protein but its P80L mutant is retained in the ER.
Sequential confocal optical sections of GOLD36 cotyledon epidermal cells expressing either GOLD36-mRFP (a) or GOLD36P80L-mRFP (a) fusion. Optical slice position is indicated on the right side of the images. (a) GOLD36-mRFP is localized in the lumen of the central vacuole (star) and complements the GOLD36 phenotype; note that ST-GFP fluorescence is mainly localized in the Golgi stacks (main image and insets). Arrowhead points to a nuclear envelope, often seen in cotyledon epidermal cells expressing ST-GFP (Faso et al., 2009). (b) GOLD36 cells expressing GOLD36P80L-mRFP show ER (arrows) and globular structures (empty arrow) due to lack of complementation. In clear contrast to GOLD36-mRFP, the GOLD36P80L-mRFP signal is in the ER but not in the vacuole (star). Insets magnification of areas of interest. Bars = 20 μm (a), 10 μm (b), 5 μm (insets).

We have also gathered evidence that GOLD36-mRFP may be localized in the ER in transient expression assays (Figures S6 and S7 and Movie S3).

Finally, we aimed to compare the fluorescence of GOLD36-mRFP with that of the GOLD36P80L mutant. Analyses of T1 GOLD36 transformants expressing GOLD36P80L-mRFP showed that, in clear contrast to the wild-type protein, the mutant protein fusion was retained in the ER network where it accumulated in bright, large structures (Figure 9b). As expected from experiments using the untagged GOLD36 (Figure 7), GOLD36P80L-mRFP did not complement the GOLD36 phenotype (Figure 9b).


It is generally believed that the morphological and functional integrity of the ER depends on the activity of ER proteins that either reside in the organelle or that are associated with its surface. Our data show that reduced export of a small vacuolar protein affects the integrity of the ER in Arabidopsis. In particular, we demonstrate that an EMS mutant has obvious displacement of membrane and soluble markers destined to post-ER compartments as well as large defects in ER network organization. Using a combination of classical positional cloning and next-generation whole genome sequencing, the mutation responsible for the phenotype was mapped to a missense mutation in the gold36 (At1g54030) locus. We established that while wild-type GOLD36 is targeted to the vacuole, the mutant GOLD36 bearing a missense P80L residue switch is retained in the ER. However, we also found that a knock-out mutant phenocopied the GOLD36 subcellular phenotype. Together these data propose the concept that a reduced or absent function of a distally targeted secretory protein can influence ER morphological and functional integrity.

GOLD36P80L is the product of an EMS allele with a unique subcellular phenotype

GOLD36 coincides with the recently described MVP1, which was identified through a screen for EMS mutants with defects in the targeting of a delta-TIP to the vacuole (Agee et al., 2010). The mvp1 allele carried a mutation that causes the missense G57E residue shift in GOLD36/MVP1; the resulting mutant showed one large membranous aggregate per cell, which, similarly to the mutant described here, contained membrane markers destined to the ER and Golgi (Agee et al., 2010). Such structures also retained membrane markers destined to post-Golgi compartments (Agee et al., 2010), indicating that similarly to GOLD36, MVP1 had defects in protein traffic at these structures. Here, we show that, differently from MVP1, GOLD36 additionally has a strong phenotype in the cortical ER with obvious retention of proteins destined to post-ER compartments in the cortical ER. Although such difference with MVP1 may be explained on the basis that the mutation in the gold36 allele encoding GOLD36P80L may be stronger compared to that of MVP1, our data combined with those of Agee et al. (2010) indicate that GOLD36/MVP1 is an important protein for ER functional and morphological integrity in plants.

GOLD36/MVP1 is a vacuolar protein

Here, we show that a GOLD36-mRFP fusion is localized at the vacuole and that the fusion is functional, complementing the mutant phenotype. Recent transient expression experiments in Arabidopsis have indicated that a GFP fusion of wild-type GOLD36 (MVP1-GFP) was localized in the ER, ER fusiform bodies, and transvacuolar strands (Agee et al., 2010), which are thin tubular structures containing cytoplasm, ER and Golgi membranes (Knebel et al., 1990; Nebenfuhr et al., 1999; Hoffmann and Nebenfuhr, 2004). However, it was not shown whether the ER-localized MVP1-GFP could complement the phenotype (Agee et al., 2010). It is possible that the ER localization of MVP1-GFP reflects the distribution of the protein pool in transit to other compartments. Lower tolerance of GFP than mRFP to acidic pH may also partially explain a lack of visualization of the marker in the vacuole (Tamura et al., 2003; Shaner et al., 2005). Furthermore, ER fusiform bodies are generally labeled by any ER lumenal protein including soluble inert ER GFP-markers (Hawes et al., 2001; Hayashi et al., 2001; Estevez et al., 2006; Faso et al., 2009) (see also Figure 3 in this work), raising the possibility that the MVP1-GFP localization in these structures may be transient. Nonetheless, the vacuolar localization of GOLD36-mRFP proposed in our work is consistent with proteomics studies that retrieved the At1g54030 product in the Arabidopsis vacuole (Carter et al., 2004), further adding to our microscopy and genetic complementation evidence that the vacuolar localization of GOLD36-mRFP reflects the distribution of the endogenous protein.

It is also worth noting that, in clear contrast to the vacuolar distribution of GOLD36, the GOLD36P80L mutant was localized in the ER. In addition, the phenotype of GOLD36/gold36P80L plants was not complemented and was similar to the knock-out mutant gold36–1. Together, these data suggest that GOLD36P80L is a non-functional protein. Secretory malfolded proteins, which may prove a hazard for the cell if delivered to distal compartments, are retained in the ER via quality control mechanisms (Brandizzi et al., 2003a; Anelli and Sitia, 2008). Therefore, the ER distribution of GOLD36P80L may be the consequence of the missense amino acid mutation in a critical residue necessary for proper folding. Alternatively, we cannot exclude that the ER retention may be linked to non-productive interactions of GOLD36P80L with ER partners that would normally facilitate ER export of GOLD36.

Although GOLD36 reaches the vacuole as a final destination, it influences ER integrity

In most eukaryotes, the membranes of the ER assume a network-like morphology characterized by the presence of interconnected cisternae as well as tubules that undergo continuous fusion and fission (Sparkes et al., 2009b). Little is known about how the complex morphology of the ER is formed and maintained, or what role the overall structure plays in the functions of the ER. It is becoming increasingly clear that ER proteins or ER-associated proteins influence the ER morphological integrity. ER tubules are pulled and extended from a membrane reservoir as cytoskeletal elements, such as microtubules in mammalian cells (Vedrenne and Hauri, 2006), or as actin filaments in plant and yeast cell polymers (Prinz et al., 2000; Brandizzi et al., 2003b; Sparkes et al., 2009a). Tubules are then stabilized by cytoskeleton-independent mechanisms; recently, ER-associated proteins including reticulons and DP1/Yop1p proteins have been implicated in such mechanisms (Voeltz et al., 2006). Similarly, ER associated dynamin-like proteins such as the mammalian atlastin and its yeast functional hortologue, Sey1p, have also been implicated in the remodeling of the ER (Hu et al., 2009; Orso et al., 2009). The ER morphology is also known to change to accommodate an overload of membrane proteins as a consequence of over-expression or defects in ER export (daSilva et al., 2004; Runions et al., 2006; Stefano et al., 2006). Our data show that GOLD36 is a vacuolar protein and that a gold36 knock-out mutant phenocopies the GOLD36 subcellular phenotype. These data show that, although GOLD36 does not accumulate in the ER at a steady state, the mutant phenotype is linked to reduced or absent GOLD36 presence in the ER. In light of this, we speculate that GOLD36 is a factor that, while in transit to its final destination, influences the integrity of the ER. This model proposes that GOLD36 participates in the maintenance of ER integrity by either working as a chaperone of other proteins that are necessary for ER integrity or by binding to and reducing the activity of proteins destined to non-ER compartments that may have deleterious effects in the ER. The model in which GOLD36 functions as a molecular chaperone is intriguing, as it may add to the repertoire of known chaperones involved in ER protein folding and maturation (Anelli and Sitia, 2008). In addition, the model would be consistent with growing evidence that ER quality control relies not only on the ERAD pathway, but also on ER chaperone-mediated protein transport of terminally malfolded proteins to the vacuole for disposal (Spear and Ng, 2003; Pimpl et al., 2006).

Here we have not pursued a secondary aspect of our work on detailing the nature of the small circular structures visualized in the ER by confocal microscopy and that of the crystalline structures contained within the GOLD36 aggregates visualized by electron microscopy. Those may be supernumerary aggregates of the GOLD36P80L or of other proteins that arise as a consequence of the GOLD36P80L mutation. We also cannot reject the possibility that such structures may be modified ER fusiform bodies. Although our electron microscopy analyses show that the GOLD36 cells contain typical ER bodies, it is possible that the GOLD36 mutation gives rise to additional ER bodies of unusual structure. Independently from the exact nature of these structures, it is tempting to speculate that they may be responsible for the anomalous appearance of the ER network.

Our genetic and microscopy analyses show that GOLD36P80L is not a dominant mutation and that the GOLD36 is phenocopied in the gold36 knock-out background. Therefore, while we cannot exclude that the presence of the mutant protein in the ER might partially exacerbate the GOLD36 phenotype in the GOLD36/gold36P80L mutant, the GOLD36P80L does not exert a dominant negative effect that would be expected if the GOLD36 structures were GOLD36P80L aggregates and solely responsible for the ER defects. Moreover, if the crystalline structures were GOLD36P80L aggregates in the ER, they should not be present in the knock-out mutant. We are currently testing this hypothesis in our lab.

It is interesting to note that the GOLD36 mutant has defects in ER protein export, as evidenced by the partial retention of a bulk flow marker in the ER. This observation leads us to ask whether GOLD36 is also required for ER protein export or whether the alteration of the ER morphology causes defects in ER export. In this model, the effect of reduced activity of GOLD36 on protein export would be indirect but it would underscore the importance of the overall integrity of the ER structure in the export functions of this organelle.

Forward genetics approach coupled with a next-generation sequencing approach for the identification of mutants of the endomembranes

Given the central position of the Golgi in the secretory pathway, a screen for Golgi integrity mutants offers the possibility to ascribe novel functions to known genes (Faso et al., 2009), but also to identify new mutants that influence the activities of other secretory organelles in relation to the Golgi (this work). Similar microscopy-based approaches to identify mutants with anomalous distribution of a particular organelle-specific marker are proving very resourceful for the identification of new players or of new mutations in known players for the integrity of the plant endomembranes (Tamura et al., 2005, 2007; Teh and Moore, 2007; Faso et al., 2009; Agee et al., 2010). Although exciting, these screens require a large investment of time and effort to isolate a sufficient number of segregating F2 individuals bearing the phenotype of interest. An alternative to the classical mapping approaches is offered by next-generation sequencers that considerably shorten the time necessary to identify the mutations responsible for aberrant phenotypes (Lister et al., 2009). In this work, we have used a combination of classical positional cloning and next-generation sequencing approaches to identify the EMS mutation responsible for the GOLD36 phenotype. Because the positional cloning required a relatively small number of segregating recessive F2 individuals with the phenotype (see Experimental Procedures section) and the sequencing was instrument-based, this approach has proven fast and cost-effective compared with standard cloning procedures. Similarly, because of the small pool of segregating F2 individuals needed for the analysis, this approach may be particularly useful for the mapping of mutants with very subtle phenotype. Therefore, we anticipate that the approach undertaken in this work may serve as a useful example to quickly identify mutations in forward genetic screens.


Here, we describe an EMS mutant that was identified through the powerful techniques of confocal microscopy and forward genetics coupled with a next-generation sequencing approach. Our results demonstrate that microscopy-based screens that rely on morphological and functional analyses of secretory organelles are useful and exciting resources for the identification of players in the integrity of organelles of the early secretory pathway. Our data add to the repertoire of factors that are known to influence the activity and the morphology of the ER and propose the concept that the activity of distally localized protein may influence the organelle’s integrity. The identification of GOLD36 is an exciting starting point toward the identification of interacting partners to establish which other factors function with this protein to influence the activities of the ER.

Experimental Procedures

Fluorescent proteins and molecular cloning

The fluorescent proteins used in this study were based on fusions with mGFP5 (Haseloff et al., 1997), EYFP (Clontech, California, http://www.clontech.com), and monomeric RFP (Campbell et al., 2002). Wild-type and mutant gold36 cDNA were amplified from cDNA of wild-type Col-0 and GOLD36 plants, respectively. The cDNA was subcloned in the binary vector pFGC5941 and expressed under the control of the CaMV 35S promoter. mRFP fusions of wild-type and mutant GOLD36 were generated by overlapping PCR of cDNA sequences (i.e. gold36 cDNAs and mRFP) followed by subcloning in pFGC5941. Constructs were confirmed by sequencing. Primer sequences used in this work are listed in Table S2.

RNA extraction and PCR analysis

RNA extraction was performed using the RNeasy Plant Mini Kit (Qiagen, http://www.qiagen.com). Reverse transcription experiments were performed using the Superscript III First Strand Synthesis Kit (Invitrogen, http://www.invitrogen.com). PCR experiments were performed in standard conditions and were carried out using 0.2 mm dNTPs, 0.2 μm primers, and 1 unit of Taq polymerase (Promega, http://www.promega.com).

Plant materials and growth conditions

We used wild-type plants of Arabidopsis thaliana (ecotypes Columbia and Landsberg erecta), a transgenic Arabidopsis line (ecotype Col-0) expressing ST-GFP. We also used the Arabidopsis mutant gold36–1 from the ABRC consortium. With the exception of the plants used for the experiment in Figures S6 and S7, plants used in this work were all stable transformants obtained with a floral dip method (Clough and Bent, 1998), and subsequent selection on MS medium supplemented with Gamborg’s B5 vitamins, 1% (w/v) sucrose, the appropriate antibiotics, and 0.8% (w/v) agar. Seeds were surface-sterilized and were grown at 21°C under 16 h light/8 h dark conditions. Cotyledons used in this work were harvested from 7-day-old seedlings. For antibiotic selection, seeds were stratified on selective medium (see above) for 5 days, then the selected seedlings were transferred to MS medium without antibiotics for 2 days prior to analysis. Selective antibiotics were glufosinate ammonium salt (BASTA; final concentration 20 μg/ml), kanamycin (100 μg/ml), or hygromycin (15 μg/ml). Transient expression was conducted with a vacuum infiltration procedure on Arabidopsis thaliana seedlings using Agrobacterium tumefaciens (OD600 = 1) carrying the binary vector for GOLD36-mRFP expression, as detailed in Marion et al. (2008).

Isolation of the GOLD36 mutant and genetic analyses

M1 and M2 ST-GFP seeds were prepared as described earlier (Faso et al., 2009). Thirty seeds from each M2 line were grown for 7 days and analyzed using confocal microscopy for displacement of the ST-GFP marker. The GOLD36 homozygous mutant was crossed with Landsberg erecta to generate a mapping population. The polymorphism between the two ecotypes was analyzed using a combination of cleaved, amplified polymorphic sequence markers and simple sequence length polymorphism markers (Konieczny and Ausubel, 1993; Bell and Ecker, 1994). The rough map was performed on 57 individuals showing the aberrant phenotype. Identification of SNPs in the rough-mapped region was achieved as follows. Genomic DNA was submitted to the Michigan State University Research Technology Support Facility for sequencing using the Illumina Genome Analyzer II (GA II) (Bentley et al., 2008). DNA was prepared and sequenced using standard kits and protocols from Illumina Inc. The final library was sequenced in seven of the eight flow cell lanes; sequencing was performed for 50 cycles. Image analysis was performed on instrument using Real Time Analysis (RTA) v 1.4. Base calling, read passing, and error analysis were performed off instrument using the Illumina GA-Pipeline v1.4. A total of 154.8 million reads (7.74 Gbp as sequence) were generated. The default CHASTITY filter of the GA-Pipeline was used for determining passed filter (PF) reads; 114.6 million reads (74%, 5.73 Gbp) passed filtering. PF reads were aligned to the Arabidopsis thaliana genome sequence (TAIR8) using the short read alignment program Bowtie (Langmead et al., 2009) (v 0.10.0). Alignment parameters were adjusted to allow up to three mismatches in the seed (the first 28 bases). The Bowtie output was converted to MAQ map format using Bowtie-Maqconvert from the Bowtie distribution. MAQ (Li et al., 2008) (v 0.7.1) was used to assemble aligned reads into a consensus to identify putative SNPs. Raw SNPs were filtered with MAQ’s SNP filter, increasing the minimum depth to call a SNP to five reads. As SNPs of interest should be homozygous, the variants were further filtered to only those in which ≥80% of the overlapping reads supported the SNP call.

Confocal laser scanning microscopy

An inverted laser scanning confocal microscope (LSM510 META, Carl Zeiss, http://www.zeiss.com) was used for confocal analyses. For GFP5 and YFP imaging, settings were as described earlier (Brandizzi et al., 2002a; Hanton et al., 2007). Imaging of DiOC6-stained cells and of GOLD36 cells labeled with propidium iodide or expressing mRFP constructs was carried out as described in Faso et al. (2009). FRAP analyses were conducted as described earlier (Brandizzi et al., 2002b). All confocal images were acquired with 1-μm pinhole settings. Post-acquisition analyses were performed with the Zeiss AIM software. PaintShop Pro was used for further image handling. Images reported in microscopy figures are representative of at least five independent experiments.

Fluorescent dyes and drug treatments

Cellular nucleic acids were stained by immersing cotyledons in a solution with propidium iodide (PI; Invitrogen; working solution: 1 μg/ml) in water for 15 min. Endomembranes were stained with the general endomembrane dye DiOC6 (working solution: 1.8 μm; Molecular Probes, http://www.invitrogen.com) in water for 30 min, as described earlier (Zheng et al., 2004). Wortmannin (working solution: 33 μm; Sigma, http://www.sigmaaldrich.com) was used on intact tissue for 12 h. All stock solutions were kept at −20°C, and working solutions were prepared fresh just before use. For analysis and observation at the microscope, samples were mounted on a slide with the solution in which they were last treated.

Electron microscopy

Cotyledons were fixed and embedded in Spurr resin as described in Faso et al. (2009), sectioned with a RMC PowerTome XL, post-stained with lead citrate, and observed with a Hitachi H-7650 transmission electron microscope (Hitachi, http://www.hitachi-hta.com). Accession numbers: gold36 (AGI: At1g54030); ubiquitin10 (At4g05320).


We acknowledge support by the Chemical Sciences, Geosciences and Biosciences Division, Office of Basic Energy Sciences, Office of Science, US Department of Energy (award number DE-FG02–91ER20021) and National Science Foundation (MCB 0841594) (F.B.) and the BBSRC (C.H.). We are grateful to Mr Kevin Carr for the bioinformatics analyses, Ms Linda Danhof for technical help, and Ms Karen Bird for editing the manuscript.