The Arabidopsis thaliana nucleotide sugar transporter GONST2 is a functional homolog of GONST1

Abstract Glycosylinositolphosphorylceramides (GIPCs) are the predominant lipid in the outer leaflet of the plasma membrane. Characterized GIPC glycosylation mutants have severe or lethal plant phenotypes. However, the function of the glycosylation is unclear. Previously, we characterized Arabidopsis thaliana GONST1 and showed that it was a nucleotide sugar transporter which provides GDP‐mannose for GIPC glycosylation. gonst1 has a severe growth phenotype, as well as a constitutive defense response. Here, we characterize a mutant in GONST1’s closest homolog, GONST2. The gonst2‐1 allele has a minor change to GIPC headgroup glycosylation. Like other reported GIPC glycosylation mutants, gonst1‐1gonst2‐1 has reduced cellulose, a cell wall polymer that is synthesized at the plasma membrane. The gonst2‐1 allele has increased resistance to a biotrophic pathogen Golovinomyces orontii but not the necrotrophic pathogen Botrytis cinerea. Expression of GONST2 under the GONST1 promoter can rescue the gonst1 phenotype, indicating that GONST2 has a similar function to GONST1 in providing GDP‐D‐Man for GIPC mannosylation.


| INTRODUC TI ON
The plant plasma membrane is an asymmetric lipid bilayer which acts as both a selective barrier and a point of contact between the interior and exterior of the cell. Glycosylinositolphosphorylceramides (GIPCs) are a glycosylated form of sphingolipid, that comprise an estimated 64% of plant sphingolipids and ~25% of the total lipids in the Arabidopsis thaliana (Arabidopsis) leaf (Bure et al. 2011;Cacas et al. 2013Cacas et al. , 2016Markham & Jaworski, 2007;Markham et al. 2006Markham et al. , 2013. GIPCs are found predominantly in the outer leaflet of the plasma membrane. GIPCs are a highly diverse class of lipids, comprising a long chain base (LCB) linked via an amide group to a fatty acid (FA) to form a ceramide, and a polar glycan head group. The diversity results from variation in the length, degree and position of unsaturation and hydroxylation of the FA and LCB, as well the structure and identity of the glycan head group. The ceramide is synthesized in the endoplasmic reticulum (ER), where it is either glucosylated to produce glucosylceramides, or it is then trafficked to the Golgi for GIPC biosynthesis.
Recent evidence supports a role for GIPC glycosylation in plantmicrobe interactions. For example, a peptide (NLP) which determines pathogenicity in many plant pathogens, including oomycetes, was shown to bind to the GIPC headgroup (Lenarcic et al. 2017). The degree of GIPC glycosylation was important in determining the degree of NLP cytotoxicity (Lenarcic et al. 2017). In Medicago, the type of GIPC glycosylation is important for the successful formation of root-microbial symbioses, both with nodulating bacteria and arbuscular mycorrhizal fungi (Moore et al. submitted), and in Arabidopsis, plants with mutated GIPC glycosylation display a constitutive hypersensitive response, including elevated salicylic acid (SA) and reactive oxygen species (ROS; Fang et al. 2016;Mortimer et al. 2013).
In addition to GIPC glycosylation, many other glycosylation reactions occur in the lumen of the Golgi, including the synthesis of polysaccharides, and glycoproteins. Nucleotide sugars are the universal sugar donors for these processes. In plants, the majority of nucleotide sugars are UDP-linked, but the GDP-linked sugars GDP-D-Mannose (GDP-Man), GDP-D-Glucose (GDP-Glc), GDP-L-Fucose (GDP-Fuc), and GDP-L-Galactose (GDP-Gal) are also critical (Bar-Peled & O'Neill, 2011). Most nucleotide sugars required in the Golgi, including all of the GDP-sugars, are synthesized in the cytosol and therefore need to be translocated into the Golgi lumen via nucleotide sugar transporters (NSTs). Many of the Arabidopsis NSTs have now been heterologously characterized (Bakker et al. 2005;Baldwin et al. 2001;Ebert et al. 2015;Handford et al. 2004;Mortimer et al. 2013;Niemann et al. 2015;Norambuena et al. 2002Norambuena et al. , 2005Rautengarten et al. ,2014Rautengarten et al. , , 2016Rautengarten et al. , , 2017Reyes et al. 2010;Rollwitz et al. 2006;Saez-Aguayo et al. 2017), although in vivo functionality is described for far fewer. Arabidopsis NSTs belong to the NST/triose phosphate translocator (TPT) superfamily which has 51 members that are distributed in six clades (Rautengarten et al. 2014). From this superfamily, only four members, the GOLGI LOCALIZED NUCLEOTIDE SUGAR TRANSPORTER (GONST) subclade, are predicted to transport GDP-sugars due to the presence of the conserved GX[L/V]NK motif (Baldwin et al. 2001;Gao et al. 2001;Handford et al. 2004).
The substrate for GMT1 is provided, at least in part, by GONST1, and indeed both gonst1 and gmt1 have very similar phenotypes (Fang et al. 2016;Mortimer et al. 2013 (Mortimer et al. 2013). However, analysis of gonst1 plants revealed a specific role in vivo as a GDP-Man transporter which provides essential substrate for GIPC glycosylation (Figure 1) (Mortimer et al. 2013). GONST2 to GONST4 were identified as GONST1 homologues on the basis of their sequence similarity to GONST1 (Handford et al. 2004). GONST4 has now been characterized as the Golgi GDP-Fuc transporter and has therefore been renamed GDP-FUCOSE TRANSPORTER1 (GFT1) (Rautengarten et al. 2016). GONST3 has recently been shown to be responsible for GDP-Gal transport and has been renamed GOLGI GDP-L-GALACTOSE TRANSPORTER1 (GGLT1) . GONST2 was also able to complement vrg4-2 (Handford et al. 2004) and able to transport all four GDP-linked sugars in vitro (Rautengarten et al. 2016), but its function in planta, as well as its specificity, remains unknown.

| Samples
All experiments were performed on at least three independently grown biological replicates unless otherwise stated.

| Plant material and growth conditions
The T-DNA line gonst2-1 (FLAG_406C01; ecotype Ws; insertion into AT1G07290) as well as gonst1-1 (FLAG_164D07; insertion into AT2G13650) were previously described in Mortimer et al. (2013). A second independent null GONST2 T-DNA insertion was not available, so two additional gonst2 alleles were generated using CRISPR/

| Subcellular localization
Agrobacterium tumefaciens (GV3101) transformed with either 35Spro:GONST2-YFP or the Golgi marker Man49-GFP (Nelson et al. 2007) were co-infiltrated into 4-week-old tobacco leaves. An additional A. tumefaciens strain carrying the p19 plasmid was also coinfiltrated to stabilize the transgene expression. Forty-eight hours after infiltration, the epidermal cells were removed from the tobacco leaves, fixed with formaldehyde and imaged using a Zeiss LSM 710 (Carl Zeiss, http://www.zeiss.com/) as previously outlined (Parsons et al. 2012).
Image analysis and processing (scale bar, brightness, and contrast) were performed using IMAGEJ (Version 1.6r) (Schneider et al. 2012).

| Histochemical detection of H 2 O 2
Detection of H 2 O 2 was by endogenous peroxidase-dependent histochemical staining using 3,3-diaminobenzidine (DAB) as described in Mortimer et al. (2013). Leaves of 15-day-old agar grown plants were submerged in 1 ml buffer (100 mM HEPES-KOH, pH 6.8) or 1 mg/ml Leaves were cleared for 30 min in 96% (v/v) ethanol solution at 70°C, and examined using a light microscope. Leaves were visually assessed as having either "no staining," "light staining," or "heavy staining."

| Quantitation of SA
For total SA determination, 500 mg leaves were frozen and ground in liquid nitrogen. The powder obtained was mixed with 1 ml 80% (v/v) methanol and incubated for 15 min at 70°C. This step was repeated four times. Pooled extracts were centrifuged and filtered through

| G. orontii infection assay
G. orontii MGH was maintained on pad4 leaves , and WT or gonst2-1 leaves were inoculated using a settling tower method, as previously described (Plotnikova et al. 1998

| GIPC analysis by TLC
Powdered, lyophilized liquid-grown callus (200 mg) was added to 5 ml of the lower layer of isopropanol:hexane:water (55:20:25) and incubated at 50°C for 15 min. Following centrifugation (500 x g, 10 min), the supernatant was transferred to a fresh tube, and the pellet was re-extracted with a further 5 ml of the lower layer of isopropanol:hexane:water (55:20:25). The supernatants were combined, dried under N2, and de-esterified by incubation with 33% (v/v) methylamine in ethanol:water (7:3) at 50°C for 1 hr. After centrifugation (500 g, 10 min), the supernatant was retained, dried under N2, and incubated in 1 ml of chloroform:ethanol:ammonia: water (10:60:6:24) overnight at 21°C with agitation. Samples were subjected to weak anion exchange chromatography as described in Mortimer et al. (2013), and following elution from the cartridge were resuspended in chloroform:methanol:[4 M ammonium hydroxide in 1.8 M ammonium acetate] (9:7:2) and separated by thin layer chromatography (TLC) using high-performance-TLC Silica gel on glass plates (Merck) developed in the same buffer. GIPCs were visualized using primuline (Skipski, 1975).

| GIPC analysis by LC/MS
Total lipid for sphingolipidomics was prepared from lyophilized tissues (5-10 mg dry weight) using a methanol/butanol-based extraction coupled with weak alkaline hydrolysis and HCl treatment to remove glycerolipids and polysaccharides, respectively, according to the previous report . Each sphingolipid species was quantified using LC-MS/MS (LCMS-8030, Shimadzu, Kyoto, Japan) with the MRM mode targeting glucosylceramides, free ceramides, and GIPCs with 0, 1, and 2 hexoses on GlcA-IPCs. The contents of Hex-GIPCs and ceramides were absolutely quantified by an internal standard-based calculation method, and GlcA-IPCs and Hex-Hex-GIPCs (for which we lack standards) were relatively quantified using the calculation factors as for Hex-GIPCs as previously described (Fang et al. 2016;Ishikawa et al. 2016).

| Cell wall monosaccharide analysis
AIR was prepared according to Mortimer et al. (2010) and 5 mg was hydrolyzed with fresh 2 M trifluoroacetic acid (TFA; 400 µl, 1 hr, 121°C). The supernatant was removed, and the pellet washed twice with water (400 µl). The supernatant and washings were combined, dried in vacuo, and analyzed by HPAEC-PAD as previously described (Fang et al. 2016

| Promoter swap
The GONST1 promoter (1.3 kb upstream of the start codon) and Constructs were transformed into Agrobacterium tumefasciens strain GV3101 and used to transform gonst1-1 with the floral dip method.
T3 plants, which were confirmed to be homozygous for the gonst1-1 T-DNA insertion (Mortimer et al. 2013), were analyzed.

| GONST2 is a close homolog of GONST1
Arabidopsis nucleotide-sugar transporters containing a con- only 19% with GFT1 (Figure 1), and is expressed at a low level in most tissues ( Figure S1). We also confirmed that GONST2 is localized to the Golgi, as previously reported (Rautengarten et al. 2016) ( Figure S2).

| Use of CRISPR to generate new gonst2 alleles
Previously, we isolated and partially characterized a homozygous gonst2-1 allele (Ws ecotype) which lacked detectable GONST2 transcript by RT-PCR but did not have a visible phenotype (Mortimer et al. 2013)). Since no further T-DNA lines were available, we used CRISPR/Cas9 gene editing to create two further gonst2 alleles, gonst2-2 and gonst2-3, in the Col-0 ecotype, as

| gonst2-1 has increased resistance to a biotrophic pathogen but not to a necrotrophic pathogen
Plant pathogens can be divided into two major groups depending on We wanted to test whether gonst1 or gonst2 plants show increased pathogen resistance, and whether this was generic or specific to biotrophic pathogens. However, gonst1-1 and gonst1-1gonst2-1 rosette leaves are not suitable for pathogen assays, as they are fully senesced by ~ 20 days under normal conditions. Therefore, we tested gonst2-1, despite the lack of a detectably significant increase in SA or ROS (Figure 2), since its rosette leaves are healthy. Indeed, gonst2-1 showed a significant increase in resistance to the biotrophic pathogen G. orontii MGH (an Arabidopsis-adapted powdery mildew) ( Figure 3), as measured by conidiophores/colony. This was in contrast to pathoassays with the necrotrophic pathogen Botrytis cinerea. Four different isolates were tested, but there was no significant difference in susceptibility between WT and gonst2-1 (Figure 3; Table 1).

| gonst2-1 and gonst1-1gonst2-1 Golgisynthesized cell wall polysaccharides are unaffected
Since GONST2 is a Golgi-localized nucleotide sugar transporter, we   (Dhugga et al. 2004;Goubet et al. 2009;Liepman et al. 2005), and it has been proposed that it has a luminal active site (Davis et al. 2010). However, no NST responsible for providing these substrates to the Golgi lumen has yet been identified. Loss of GONST1 does not affect glucomannan biosynthesis (Mortimer et al. 2013), GFT1 is a GDP-Fuc transporter (Rautengarten et al. 2016), and GGLT1 is a GDP-Gal transporter ).
Since glucomannan is a relatively minor component of the cell wall (Handford et al. 2003), the monosaccharide analysis may not reveal alterations to its quantity or the Glc:Man ratio of the glucomannan backbone. Therefore, we used Polysaccharide Analysis by Carbohydrate gel Electrophoresis (PACE) to investigate glucomannan quantity and structure (Handford et al. 2003). No difference was seen either in the type or quantity of the oligosaccharides released by hydrolysis of gonst2-1, gonst1-1gonst2-1, and WT AIR by mannanases ( Figure 5). Therefore, this suggests that GONST1 and GONST2 are not providing substrate in the Golgi lumen for mannan biosynthesis.

| gonst1-1gonst2-1 has less cellulose
Previously, we showed that a mutant in GIPC mannosylation (GMT1) has reduced cellulose (Fang et al. 2016). To test whether this phenotype is common to plants with altered GIPC mannosylation, we hydrolyzed the TFA-insoluble AIR fraction with sulfuric acid to release glucose derived from cellulose ( Figure 6). gonst1-1 had a significant reduction in upper and lower stem cellulose content as compared to WT, whereas callus and seedling were unaffected.
gonst2-1 did not show a significant difference in any tissue type analyzed compared to WT. However, gonst1-1gonst2-1 mutants showed a significant decrease in callus cellulose content (a tissue rich in primary cell wall), compared to the WT or single mutants, but not in other tissue types. These data are consistent with a specific role for GIPC mannosylation in determining cell wall cellulose content.

| Expression of GONST1pro:GONST2 in gonst1-1 rescues growth and GIPC glycosylation
To test whether the phenotypic differences observed between gonst1 and gonst2 are due to differences in the protein function, or whether they are due to differences in expression level, we expressed the GONST2 coding sequence (CDS) driven by either the GONST1 promoter (GONST1pro:GONST2) or the GONST2 promoter (GONST2pro:GONST2) in the gonst1-1 background. Multiple independently transformed lines were selected for analysis. Analysis of T3 segregants revealed that some of the homozygous gonst1-1 plants had a restored growth phenotype (Figure 7). GONST2 expression was analyzed by real-time RT-PCR (Figure 7). The suppression of the gonst1-1 growth phenotype was only apparent in those lines in which GONST2 expression was driven by the GONST1 promoter (Figure 7). The rescue of the growth phenotype was reflected in the biochemical characterization of GIPC headgroup composition (Figure 7). This result supports the view that GONST2 has the same function as GONST1, and that the function is cell type specific and/or dose dependent.

F I G U R E 3
Susceptibility of gonst2-1 plants to biotrophic and necrotrophic pathogens. (a) 5 days after inoculation with the biotroph G. orontii, leaves were harvested, stained with trypan blue and conidiophores per colony counted. The data represent the mean of 12-30 leaves per genotype per experiment, scored in three independent experiments, ±SD (Student's t-test, * p <.05, *** p <.001

| D ISCUSS I ON
The aim of this research was to characterize the role of the final bona fide member of the GONST clade of four GDP-sugar transporters, GONST2. Our data support the conclusion that, while in vitro it has been reported to transport all GDP-linked sugars (Rautengarten et al. 2016), in planta it has a specific role in providing GDP-Man for GIPC glycosylation. We also show that GONST2 is a functional homolog of GONST1. Since gonst2 does not display the severe growth defects of gonst1, we were able to perform pathoassays to investigate the previously reported constitutive defense response of gonst1 (Mortimer et al. 2013), and show that gonst2-1 has increased resistance to the biotrophic pathogen G.
orontii but not the necrotrophic pathogen B. cinerea. Note that these assays were only performed on a single allele of gonst2, and so remain to be confirmed using the additional gonst2 alleles or complementation lines.
Resistance to biotrophic pathogens, such as the powdery mildew-causing G. orontii, is regulated by SA signaling (Wildermuth et al. 2001 The haustorium is surrounded by host-derived membrane called the extrahaustorial membrane, which has modified endosomal characteristics (Inada, Betsuyaku, et al., 2016). It has been shown that GIPCs are important for secretory sorting of proteins (Markham et al. 2011;Wattelet-Boyer et al. 2016), and therefore, it may be that changes to a minor class of GPIC are enough to disrupt these pro-

F I G U R E 4 Glycan headgroup composition of gonst GIPCs (a)
TLC of a GIPC-enriched membrane fraction which has been stained with primuline. Bands discussed in the text are marked with a red arrow head. (b) An enriched GIPC fraction was analyzed by LC-MS/ MS MRM. The data here are collapsed to describe only the number of hexoses on the GIPC headgroup. All data are mean ± SD of three independently grown replicates of liquid grown cell culture; asterisk indicates significant difference from the wild type (Student's t-test, *** p <.001). The full dataset is shown in Figure S5 and Dataset S1 phenotype was also reported for gmt1, which has the same biochemical GIPC phenotype as gonst1gonst2 (Fang et al. 2016). The reasons for this decrease are not clear. It is possible that the altered GIPC glycosylation affects trafficking of the rosettes to the plasma membrane, or alternatively, the change to plasma membrane composition affects CESA function. CESA proteins are Sacylated, and it has been suggested that this decoration may either localize proteins to lipid microdomains (which are rich in GIPCs) or even facilitate their formation (Konrad & Ott, 2015;Kumar et al. 2016). COBRA and COBRA-like proteins which are also essential for normal cellulose biosynthesis are glycosylinositolphosphatidylinositol (GPI) anchored (Roudier et al. 2005). GPI anchored proteins are targeted to the outer leaflet of the plasma membrane, and to lipid microdomains . Therefore, correct GIPC glycosylation may be necessary for either CESA activity or localization and retention of GPI-anchored proteins in the plasma membrane. More recently, a role for GIPCs in modulating the salt-dependent activation of a plasma membrane calcium channel (Jiang et al. 2019) and for plasmodesmata function (Yan et al. 2019) suggesting that GIPCs may have a broad role in regulating plasma membrane functionality. Mannan content is unchanged in the gonst1gonst2 plants. It had been reported that CSLA9, unlike related GT2 proteins (CSLC4, CESAs), has a topology which results in a luminal active site (Davis et al. 2010). This would necessarily require a nucleotide transporter to provide GDP-sugars for mannan biosynthesis. However, none of the predicted GDP-sugar transporters seem to have this function in planta (Mortimer et al. 2013;Rautengarten et al. 2016;Sechet et al. 2018). This implies that either the mannan synthases do not require a nucleotide sugar transporter, or that the transporter does not have a canonical GDP-binding motif.
Future work will be required to confirm these data, making use of the additional gonst2 alleles now available (Liang et al. 2019). It will also be important to establish how GIPC glycosylation affects these assorted membrane-based processes. For example, molecular dynamics could be applied to model the plant plasma membrane and understand how the GIPC glycan headgroup structure affects protein movement within the membrane. It will also be interesting to understand what drives the differences in functionality of the NSTs in in vitro assays versus in planta function. Both GONST1 and GONST2 can transport all GDP-sugars when tested in liposomebased assays (Mortimer et al. 2013;Rautengarten et al. 2016), but it is clear that they are highly specific in vivo. This could be mediated by substrate concentration, interaction with non-catalytic proteins, or interactions with the GT that utilizes the substrate (in this case GMT1 (Fang et al. 2016)). The recent crystal structure of the yeast Vrg4 NST provided new insights into how NST function is regulated F I G U R E 5 PACE fingerprint of mannan in stem cell walls of WT, gonst1, gonst2, and gonst1gonst2. Oligosaccharides released from AIR by mannanase digestion were derivatized with 8-aminonapthalene-1,3,6-trisulfonic acid (ANTS) and visualized by PACE. (Man) 1-6 oligosaccharides were used as a standard. A representative gel from multiple experiments is shown F I G U R E 6 gonst1gonst2 has reduced crystalline cellulose. Crystalline cellulose content was determined as the glucose content released by sulfuric acid treatment of the TFA-insoluble fraction of AIR. All data are mean ± SD of three biological replicates. Asterisk indicates significant difference from the wild type (Student's t-test, ** p <.01) (Parker & Newstead, 2017). To our knowledge, no plant NSTs have yet been structurally characterized, but we expect that this information will be critical for understanding NST specificity.

ACK N OWLED G M ENTS
The authors would like to thank Professor Taku Demura, Dr Misato Ohtani, and the rest of the Demura research team for their support at RIKEN.

CO N FLI C T S O F I NTE R E S T
No conflicts of interest to declare. T-2, Fusion2 T-3, and Fusion2 T-4 seedlings. Scale bar = 1 cm. Bottom row: 6-week-old WT, gonst1-1, Fusion1 T-1, Fusion1 T-2, Fusion2 T-3, and Fusion2 T-4. Plants were first grown on agar for 10 days, and then transplanted onto soil. Bar = 3 cm. (c) Gene expression analysis of GONST2 relative to WT Ws and normalized against TUBULIN using Q-PCR. Values represent average of three biological replicates ±SD. (d) An enriched GIPC fraction was analyzed by LC-MS/MS MRM. The data here are collapsed to describe only the number of hexoses on the GIPC headgroup. All data are mean ± SD of three independently grown replicates of liquid grown cell culture. Asterisk indicates significant difference from the wild type (Student's t-test, * p <.05, *** p <.001)