A Missense Mutation in the NSF Gene Causes Abnormal Golgi Morphology in Arabidopsis thaliana

. The Golgi apparatus is a key station of glycosylation and membrane traffic. It consists of stacked cisternae in most eukaryotes. However, the mechanisms how the Golgi stacks are formed and maintained are still obscure. The model plant Arabidopsis thaliana provides a nice system to observe Golgi structures by light microscopy, because the Golgi in A. thaliana is in the form of mini-stacks that are distributed throughout the cytoplasm. To obtain a clue to understand the molecular basis of Golgi morphology, we took a forward-genetic approach to isolate A. thaliana mutants that show abnormal structures of the Golgi under a confocal microscope. In the present report, we describe characterization of one of such mutants, named #46-3. The #46-3 mutant showed pleiotropic Golgi phenotypes. The Golgi size was in majority smaller than the wild type, but varied from very small ones, sometimes without clear association of cis and trans cisternae, to abnormally large ones under a confocal microscope. At the ultrastructual level by electron microscopy, queer-shaped large Golgi stacks were occasionally observed. By positional mapping, genome sequencing, and complementation and allelism tests, we linked the mutant phenotype to the missense mutation D374N in the NSF gene, encoding the N -ethylmaleimide-sensitive factor (NSF), a key component of membrane fusion. This residue is near the ATP-binding site of NSF, which is very well conserved in eukaryotes, suggesting that the biochemical function of NSF is important for maintaining the normal morphology of the Golgi.


Introduction
The Golgi apparatus exists in all eukaryotic cells, which was first identified by Camillo Golgi. This organelle plays crucial roles in cargo sorting and processing such as glycosylation. The Golgi apparatus usually consists of flattened disk-shaped membrane called cisternae, each of which has distinct processing enzymes. These enzyme distributions show a conserved cis-to-trans polarity that reflects the order of oligosaccharide processing (Emr et al., 2009), thus early acting enzymes dominate in cis-cisternae, whereas late-acting enzymes are concentrated in trans-cisternae (Nilsson et al., 2009). The cis-side cisternae receive newly synthesized proteins from the endoplasmic reticulum (ER), which travel across the stack to the trans-side cisternae. Then they are sorted in the trans-Golgi network (TGN) and transported to their final destinations such as outside of the cell, the plasma membrane (PM) and lysosomes/vacuoles.
There has been a debate about how cargo proteins are transported in the Golgi. From mammalian-cell-based cell free assays, the vesicular transport model has been proposed, which assumes that the Golgi cisternae are stable static compartments and resident Golgi proteins are retained in the cisternae, while cargo proteins are conveyed by vesicles in the anterograde way (Rothman and Wieland, 1996). Another model, called the cisternal maturation model, considers that resident Golgi proteins move from later to earlier cisternae, thereby cisternae mature from cis to trans. Cargo can just stay in cisternae and eventually reach the trans side. This model was strongly supported by live imaging in the budding yeast Saccharomyces cerevisiae (Losev et al., 2006;Matsuura-Tokita et al., 2006), which clearly demonstrated that Golgi cisternae change their properties from cis to trans over time. In either of the two models, vesicles are believed to play critical roles in transport. General mechanisms of vesicular transport should apply to these intra-Golgi events. Namely, vesicles are formed by the budding machinery, which comprises COPI (coat protein complex I) and the Arf GTPase, and are targeted to a particular cisterna by the fusion machinery, which involves tethers, Rab GTPases, NSF (Nethylmaleimide-sensitive factor), SNAP (soluble NSFattachment protein) and SNAREs (SNAP receptors). Indeed, the cisternal maturation in yeast is strictly controlled by COPI, again supporting the main frame of the model (Ishii et al., 2016;Papanikou et al., 2015). It should also be noticed, however, that tubular extensions and interconnections can also play roles in intra-Golgi transport (Glick and Luini, 2011;Glick and Nakano, 2009).
The Golgi consists of stacked cisternae in most eukaryotes. This principal structural feature of the Golgi is considered to reflect its function (Mollenhauer and Morré, 1991). However, there are marked variations among organisms in the morphological and organizational patterns of the Golgi. In the budding yeast S. cerevisiae, the Golgi does not form stacks and individual cisternae as transient structures are dispersed in the cytoplasm (Losev et al., 2006;Matsuura-Tokita et al., 2006;Preuss et al., 1992). In mammalian cells by contrast, the Golgi forms a giant and complex ribbon-shaped structure, which is densely concentrated to the perinuclear centrosomal region. In flowering plants such as Arabidopsis and tobacco, the Golgi is in the form of mini-stacks of cisternae, which are individually functional units, being distributed throughout the cytoplasm and moving along actin filaments. In some green algae, the Golgi represents a large stack consisting of numerous cisternae (Noguchi, 1978;Staehelin and Kang, 2008). The Golgi apparatus in flowering plants is thus regarded as a nice model to study its morphology in detail by light microscopy.
To obtain a clue to understand the molecular mechanisms how the Golgi stacks are formed and maintained, we employed a forward-genetic approach combined with confocal microscopy-based visualization to isolate mutants from A. thaliana that show abnormal morphology of the Golgi. Among 30 mutant candidates we isolated, we describe in the present paper about the mutant #46-3, which shows aberrant structures of the stack. By positional mapping, genome sequencing, and complementation and allelism tests, we concluded that a D-to-N missense mutation in the NSF gene, encoding a key factor in membrane fusion, is the cause of the phenotype.

Plant materials and growth conditions
For wild-type plants of A. thaliana, ecotypes Colombia-0 (Col-0) and Landsberg erecta were used in this study. Seeds of the transgenic plant of A. thaliana (Col-0) expressing ERD2-GFP (Boevink et al., 1998;Takeuchi et al., 2000Takeuchi et al., , 2002 (named A21) were mutagenized by treatment with 0.3% ethyl methanesulfonate (EMS) for 16 h. T-DNA insertion mutants, SALK_138721, SAIL_1155_C06 and SAIL_620_E12 were obtained from the Arabidopsis Biological Resource Center (Columbus, OH, USA). Mutants were backcrossed three times with the wild-type Col-0. Surviving seedlings were individually grown up and subjected to self-pollination to establish M2-generation lines. M2 seeds were sown on MS medium [1×MS salt, 1% sucrose, vitamin mix and 0.2% agar], vernalized in the dark at 4°C for 4 days, and grown at 23°C under continuous light. Selection media contained glufosinate ammonium salt (BASTA; final concentration 7.5 μg/ml) or kanamycin (50 μg/ml) for T-DNA insertion lines and hygromycin (25 μg/ml) for complementation lines.

Isolation of the #46-3 mutant and genetic analysis
M2 ERD2-GFP seeds were grown for 16 days and observed under a confocal microscope. A homozygous mutant line named #46-3, showing abnormal morphology of the Golgi, was crossed with Landsberg erecta to generate a mapping population. A. thaliana ecotype-specific markers of simple sequence length polymorphisms (SSLPs) and cleaved amplified polymorphic sequences (CAPSs) (Konieczny and Ausubel, 1993;Bell and Ecker, 1994) were used for rough mapping on 20 individuals showing the abnormal Golgi phenotype. For genome-sequencing and SNPs calling, a modified method from Uchida et al. (2011)

Confocal microscopy
For single-color imaging, transgenic plants were visualized under an Olympus IX81 fluorescence microscope equipped with a confocal laser-scanning unit (CSU10, Yokogawa Electronic, Tokyo, Japan) and images were acquired by a CCD camera ORCA-R2 (Hamamatsu Photonics, Hamamatsu, Japan). Dual-color imaging was carried out with a LSM780 confocal microscope (Zeiss, Jena, Germany). The central regions of petioles 16 days after germination were mounted with water on glass slides. Images were processed and analyzed with ImageJ 1.49i (National Institute of Health, Bethesda, MD, USA).

Electron microscopy
16-day-old petioles of seedlings (wild type and #46-3) were rapidly frozen in a high-pressure freezer (HPM010; Bal-Tec). Frozen samples were transferred to 2% osmium tetraoxide in anhydrous acetone that had been precooled with liquid nitrogen. Samples were maintained at -80°C for 7 days, at -20°C for 2 h, at 4°C for 2 h, and then at room temperature for 2 h. After several washes with anhydrous acetone, samples were embedded in Spurr's resin (Nisshin EM). Ultrathin sections (thickness 60-80 nm) were cut, stained with uranyl acetate and lead citrate, and observed under a transmission electron microscope (JEM-1010; JEOL).

Isolation of a mutant with abnormality in the Golgi morphology
To isolate mutants that show abnormalities in the morphology of the Golgi apparatus, we decided to use the A. thaliana line that expresses ERD2-GFP driven by 35S promoter, a well-established cis-Golgi marker (Boevink et al., 1998;Takeuchi et al., 2000Takeuchi et al., , 2002. The fluorescence of ERD2-GFP shows disk-like structures, which are dispersed through the cytoplasm. The parent line A21, expressing ERD2-GFP, was treated with ethyl methanesulfonate to induce mutations. After due processes, M2 lines were established and subjected to visual screening under a confocal microscope, Olympus IX81 equipped with a spinningdisk confocal unit CSU10. As the morphology of the Golgi appeared to vary in different tissues, we selected the thirdleaf-petioles of 16-day-old seedlings, which reproducibly show large and bright Golgi structures. By manual examination of approximately 10,000 M2 lines, we selected 30 mutant candidates that showed altered shapes of the Golgi. The morphological phenotypes were of a wide variety; large, small, with tubules, large and bent, large aggregates, and so on. They were further subjected to crossing with the parent to determine dominant or recessive natures. The mutant line named #46-3, which is the subject of the present paper, displayed abnormalities in the size and shape of the Golgi and this phenotype turned out recessive to the wild type (see below). Otherwise, the #46-3 mutant did not show any discernible macroscopic phenotypes. The mutant plants grew normally and were fertile.
The abnormality of the Golgi morphology in #46-3 was the peculiarity of size and shape as shown in Fig. 1A. While the ERD-GFP fluorescence in the wild type showed mostly disk-like shapes in similar sizes, in the #46-3 mutant the shape looked irregular and the size varied from smaller to larger as compared to the wild type ( Fig. 1A and B). To quantify the Golgi size, we measured the areas of fluorescent signals in 10 epidermal cells of petioles (expressed as pixels). The distribution of the Golgi size in the wild type and in #46-3 is shown as histograms in Fig. 1C. It clearly shows that the Golgi size of #46-3 is in majority smaller than the wild type. It should be noted that abnormally large Golgi was also occasionally seen in the #46-3 mutant (Fig.  1D).

Localization analysis of cis-and trans-Golgi in the #46-3 mutant
To further investigate the anomalies of the Golgi morphology, the arrangement of cis and trans cisternae in Golgi stacks was examined in the wild-type and #46-3 plants stably expressing ST-mRFP, a trans-Golgi marker (sialyl transferase from rat) (Boevink et al., 1998;Ito et al., 2012Ito et al., , 2017, in addition to ERD2-GFP. As shown in Fig. 2A, fluorescence signals of ERD2-GFP and ST-mRFP were mostly adjacent to each other. Sometimes the distance between the peaks of GFP and mRFP signals was markedly larger in the #46-3 mutant (Fig. 2B). GFP signals without associating mRFP signal were also occasionally observed in the #46-3 mutant (Fig. 2C).

Electron microscopic analysis of the Golgi structure in the #46-3 mutant
We next analyzed the Golgi morphology in the #46-3 mutant by transmission electron microscopy. The Golgi stacks in the wild-type Arabidopsis usually consist of around 5 cisternae as shown in Fig. 3A. Abnormally small Golgi structures, which were often seen in fluorescence, were not easy to identify in the leaf epidermal cells of #46-3, perhaps because they did not show typical stacks. Fig. 3B shows one example of electron micrograph showing two Golgi-like structures harboring 2-3 cisternae. We sometimes observed abnormal Golgi stacks consisting of increased numbers of cisternae in the #46-3 mutant. In the example shown in Fig. 3C, as many as 9 cisternae can be counted in what appeared to be a single Golgi stack. In another example (Fig. 3D), a Golgi stack bore an unusually large trans-most cisterna. To quantify the cisternal number per Golgi stack, we counted that in the epidermal cells of petioles. The distribution of the cisternal number in the  wild type and #46-3 is shown as histograms in Fig. 3E. The cisternal numbers did not appear significantly different in the Golgi stacks between the wild type and the #46-3 mutant, the latter showing a broader distribution. As mentioned above, abnormal Golgi that did not show typical stacks might be overseen in electron micrographs. Combined with the fluorescence observation, we concluded that the #46-3 mutant had a pleiotropic defect in normal arrangement of Golgi stacks in both size and shape.

The #46-3 Golgi phenotype is linked to a mutation in the At4g04910 locus
To identify the mutation in the genome that caused the phenotype of #46-3, we proceeded to identify the mutation locus. We crossed the #46-3 plants with the wild-type plants (Col-0) expressing ERD2-GFP to determine the inheritance pattern. Segregation of the F2 population from the third backcross indicated that the #46-3 mutation was recessive to the wild type. To generate a mapping population, we crossed #46-3 with a different ecotype Landsberg erecta. The genomic DNA of F2 plants exhibiting the abnormal phenotype was roughly mapped with simple sequence length polymorphism and cleaved amplified polymorphic sequence markers. Then, the candidate region containing the mutation was analyzed by the linkage analysis using a next-generation DNA sequencer. The results of these analyses nailed down the causal mutation on the upper arm of the chromosome 4 at 1-5.5 Mbp interval. Among nucleotide differences within this region, only 3 were predicted to be non-silent in open-reading frames. They were in the loci At4g02410, At4g02750, and At4g04910, which were annotated to encode L-type lectinlike protein kinase 1 (AtLPK1), tetratricopeptide repeat (TPR)-like superfamily protein, and N-ethylmaleimidesensitive factor (NSF), respectively. Among these three, At4g04910 was of particular interest to us, because NSF is well known as a key protein required for membrane fusion. The At4g04910 locus of #46-3 had a G-to-A point mutation at the nucleotide number 2492290, which led to a GAT to AAT codon change resulting in the amino acid change from aspartic acid to asparagine at the residue position 374 (D374N). A. thaliana has only one ortholog of NSF in the genome. Comparison of NSF sequences indicates that this aspartate residue is very well conserved in eukaryotes (Fig.  4).

The missense mutation in NSF is responsible for the Golgi phenotype of #46-3
To examine whether the D374 mutation in NSF was indeed responsible for the observed phenotype, we performed a complementation test. We cloned the genomic DNA fragment from the wild-type plants containing the whole At4g04910 locus including the promoter and the openreading frame and introduced into the #46-3 mutant. We analyzed the Golgi morphology in leaf petioles of two independent T2 seedlings under a confocal microscope. As shown in Fig. 5, the #46-3 phenotype in the Golgi shape (Fig. 5A) and the Golgi size distribution (Fig. 5B) was almost fully complemented. DNA fragments containing At4g02410 or At4g02750 did not complement the defect.
We also conducted an allelism test. We obtained from the ABRC stock center three A. thaliana lines containing T-DNA insertions at the NSF gene locus in the Col-0 background. We confirmed that T-DNA was inserted in the 5' UTR both in SALK_091598/SALK_138721 (at nucleotide 2495839) and in SAIL_1155_C06 (at nucleotide 2495824) and in the third exon in SAIL_620_E12 (at nucleotide 2495008) of the NSF gene. We could make a homozygous line of SALK_091598/SALK_138721 but not for SAIL_1155_C06 and SAIL_620_E12, probably because the complete knockout of NSF is lethal. Then, we crossed #46-3 with SAIL_1155_C06 to establish a heterozygous line and examined its Golgi morphology. As shown in Fig.  6, the Golgi phenotype seen by ERD2-GFP signals in the transformed #46-3/SAIL_1155_C06 plant phenocopied that of the #46-3 mutant under a confocal microscope (Fig. 6A).
Taken together, we concluded that the D374 missense mutation in NSF caused the Golgi morphology phenotype of #46-3.

Discussion
Morphology of organelles is considered to reflect their functions. Stacking of cisternae in the Golgi apparatus is one of the most striking structural features, which has been attracting many cell biologists, and is believed very important for efficient processing and sorting of cargo molecules. However, the fact that the budding yeast S. cerevisiae does not have a stacked structure of the Golgi yet managing efficient cargo processing leaves the significance of stacking elusive. From a mechanistic point of view, how these stacked structures are formed and maintained is also intriguing. The roles of Golgi matrix proteins have been argued for animal cells, but most of these molecules are not conserved in plants, whose Golgi nevertheless shows beautiful stacked structures. We have recently revealed by superresolution confocal live imaging microscopy (SCLIM) (Kurokawa et al., 2013) that Golgi stacks in tobacco cells NSF Mutation Causing Abnormal Golgi Morphology are formed in the order from cis-to trans-cisternae and that the presence of scaffold at cis-most side, which we named Golgi entry core compartment (GECCO), plays a role of receiving cargo from the ER (Ito et al., 2012(Ito et al., , 2017. Furthermore, computational simulation has implicated selforganizing properties of Golgi cisternae during reassembly processes (Tachikawa and Mochizuki, 2017).
To obtain further insights into the molecular basis for Golgi morphogenesis in plant cells, we have sought for mutations that affect Golgi morphology in Arabidopsis by a forward-genetic approach. In the present paper, we have demonstrated that the mutant #46-3 has a D374N missense mutation in NSF, which is responsible for the morphological defect of the Golgi.

NSF is a key player in membrane fusion events
NSF, N-ethylmaleimide-sensitive factor, was first identified as a key protein functioning in intra-Golgi trafficking. Inactivation of the Golgi membrane by N-ethylmaleimide in a cell-free reconstitution system caused perturbation of intra-Golgi transport (Glick and Rothman, 1987). This assay allowed purification of NSF (Block et al., 1988). As interacting molecules with NSF, SNAP was isolated (Weidman et al., 1989), and as the membrane receptors for SNAP, SNARE proteins were discovered (Söllner et al., 1993). All these components are now known to constitute the pivotal machinery for membrane fusion.
NSF is a member of the ATPases associated with diverse cellular activities plus (AAA+) family, which are present in all kingdoms of eukaryotes (Ryu et al., 2016;Zhao and Brunger, 2016). The most important role of ATP hydrolysis by NSF is to disassemble the SNARE complex after membrane fusion. NSF is now known to be essential for numerous membrane fusion events (Südhof, 2013;Sutton et al., 1998;Weber et al., 1998).
As such, the NSF function is essential for many cellular activities and thus must be essential for life. Indeed, the S. cerevisiae SEC18 gene that encodes NSF is essential for growth. Our results that A. thaliana lines that were homozygously null for NSF were never obtained from two T-DNA insertion lines (in this study), probably indicate that the complete loss of NSF is lethal.
The missense mutation of NSF that we identified as the cause of the abnormal Golgi morphology was D to N substitution at the amino acid residue #374. A D-to-N mutation is often regarded as subtle because the structural change in the side chain is small. It is probably why this mutant was obtained without an appreciable growth phenotype. Microscopically, the phenotype of this mutant is pleiotropic. At the fluorescence confocal microscope level, the Golgi size was in majority smaller than the wild type, but varies from very small ones, sometimes without clear association between cis and trans cisternae, to abnormally large ones, which were rather rare. At the ultrastructural level by transmission electron microscopy, abnormally small Golgi structures were hard to find, perhaps because they did not look like typical Golgi, but queer-shaped large Golgi stacks were occasionally observed.

Asp374 in the NSF-D1 domain is highly conserved among eukaryotes
In the process of membrane fusion, several reactions proceed in a sequential manner. The leading player is the SNARE proteins; R-SNARE on the vesicle membrane and the Qa-, Qb-and Qc-SNAREs on the target membrane, form 4-helix bundles to execute physical fusion of two lipid bilayers. After fusion, the SNARE bundles have to be disassembled for the next round of fusion reactions. NSF and SNAP play roles in this SNARE disassembly. Hydrolysis of ATP by NSF is essential for the disassembly reaction.
The molecular structure of NSF has been analyzed by Xray crystallography (Lenzen et al., 1998;May et al., 1999). NSF consists of an amino-terminal region that interacts with other components of the vesicle trafficking machinery, followed by the two homologous ATP-binding cassettes, designated D1 and D2, which possess essential ATPase and hexamerization activities, respectively. The D374 residue of Arabidopsis NSF lies in the D1 domain near the ATPbinding site, around which the amino acid sequence is very well conserved among eukaryotes (Fig. 4). Thus the D374N mutation of Arabidopsis NSF may well affect its ATPbinding or hydrolysis activities of NSF, which should be examined in detail on purified proteins in the future.

Membrane fusion and the Golgi morphology
Now the question is how the lesion of NSF can affect the Golgi structure. Is it the membrane fusion events that are directly involved in cisternal arrangements? Regardless the model of cargo transport within the Golgi, it is obvious that the membranes in the Golgi have to be in a very dynamic equilibrium. Vesicle budding and membrane tubulation together with membrane fusion events must be always going on to maintain the steady-state morphology of the Golgi. Finding of a missense mutation in NSF, the essential factor of fusion, will provide us a clue to further our understanding of how the beautiful structure of the Golgi is formed and maintained.