N-glycosylation is crucial for trafficking and stability of SLC3A2 (CD98)

The type II glycoprotein CD98 (SLC3A2) is a membrane protein with pleiotropic roles in cells, ranging from modulation of inflammatory processes, host–pathogen interactions to association with membrane transporters of the SLC7 family. The recent resolution of CD98 structure in complex with LAT1 showed that four Asn residues, N365, N381, N424, N506, harbour N-glycosylation moieties. Then, the role of N-glycosylation on CD98 trafficking and stability was investigated by combining bioinformatics, site-directed mutagenesis and cell biology approach. Single, double, triple and quadruple mutants of the four Asn exhibited altered electrophoretic mobility, with apparent molecular masses from 95 to 70 kDa. The quadruple mutant displayed a single band of 70 kDa corresponding to the unglycosylated protein. The presence in the membrane and the trafficking of CD98 were evaluated by a biotinylation assay and a brefeldin assay, respectively. Taken together, the results highlighted that the quadruple mutation severely impaired both the stability and the trafficking of CD98 to the plasma membrane. The decreased presence of CD98 at the plasma membrane, correlated with a lower presence of LAT1 (SLC7A5) and its transport activity. This finding opens new perspectives for human therapy. Indeed, the inhibition of CD98 trafficking would act synergistically with LAT1 inhibitors that are under clinical trial for anticancer therapy.


LAT1
Large neutral amino acids transporter small subunit 1 LAT2 Large neutral amino acids transporter small subunit 2 xCT Cystine/glutamate transporter PNGase N-Glycosidase F HEK293 Human embryonic kidney 293 cells DMEM Dulbecco's modified eagle medium Tris Tris(hydroxymethyl)aminomethane FBS Fetal bovine serum EDTA Ethylenediaminetetraacetic acid TX-100 T-Octylphenoxypolyethoxyethanol PBS Phosphate buffered saline BCH 2-Amino-2-norbornanecarboxylic acid The protein CD98, also known as 4F2hc, is a type II glycoprotein formed by an N-terminus intracellular domain, one transmembrane helix, and a large hydrophilic C-terminus portion (Fig. 1). CD98 is ubiquitously expressed in normal tissues and, since the very first studies, it has been found in several tumour cell lines 1 . Even though "CD98" and "4F2hc" are commonly used names, the protein was later included in the third family of the SoL-uteCarrier superfamily upon the resolution of the human genome and the evolutionary classification of SLCs.
In particular, CD98 is the SLC3A2 member of the family, together with rBAT (SLC3A1). This puzzling naming and classification reflect the sizable number of biological functions identified and characterized over the years

Results
The presence of N-glycosylated residues in CD98 was deduced based on structural observations and in silico analyses conducted with bioinformatic tools. As shown in Fig. 1, CD98 possesses four Asn residues namely N365, N381, N424, N506 that are predicted as N-glycosylation sites by NetNGlyc-1.0 32 ; indeed, the four residues carry glycosyl moieties as demonstrated also by the recently solved LAT1/CD98 complex structure 20,21 . The mentioned Asn residues are located at the surface of the protein, protruding towards the extracellular matrix. The presence of N-glycosylated residues correlates well with the higher apparent molecular mass with respect to the theoretical mass of the sole protein moiety of CD98 (Fig. 2). To investigate the possible role of the N-glycosyl moieties in the trafficking towards the cell membrane, we produced site-directed mutants of CD98 in which each of the N residues, carrying the glycosyl moieties, was substituted with Q. Each mutant was tagged at the N-terminus with FLAG. Moreover, to obtain a glycosyl-free protein, the quadruple mutant (Qm) was constructed by passing through double (Dm: N381/424Q) and triple (Tm: N365/381/424Q) mutants. From the WB analysis, it appears that the loss of a single glycosylated site causes only small changes in the apparent molecular mass of CD98, whereas Dm, Tm and Qm mutants showed a more evident shift of the apparent molecular mass compared to WT ( Fig. 2 upper panel). After treatment with PNGase, which specifically removes all the N-linked glycosyl moieties, both WT and mutants showed the same molecular mass of 70 KDa, close to the theoretical mass of the CD98 protein moiety. This finding confirmed that the four identified N residues are the only competent in N-glycosylation of hCD98 ( Fig. 2 lower panel), in agreement with the prediction (Fig. 1) and the 3D structure data 20,21 . The mutant constructs were transiently transfected into HEK293 cells to assess the expression of hCD98 at the cell membrane by confocal microscopy using two different antibodies for detecting the total over-expressed protein (anti-FLAG) or the membrane localised CD98 (anti-CD98). The detection with anti-CD98 was used as an internal control of the membrane surface since the antibody epitope is extracellular (Fig. 1). As shown in Fig. 3A, each of the single mutants did not show any appreciable difference in membrane localization compared to the WT. Among the multiple mutants (Fig. 3B), the Dm (N381/424Q) and the Tm (N365/381/424Q) exhibited some differences with respect to the WT. The Qm showed the most evident difference in membrane localization compared to the WT (Fig. 3B). In particular, the anti-FLAG images (red) showed several protein spots in intracellular sites that were not present in the anti-CD98 images (green). The merge of the two fluorescence signals highlighted the presence of the non-merged spots in the endocellular area (Fig. 3B). To confirm the data from confocal microscopy, a biotinylation assay was performed. The molecular mass of the Qm correlated well with the lack of glycosyl moieties of this protein (Fig. 4A). A clear decrease in the amount of Qm present in the membrane was observed with respect to the total protein (Fig. 4A, lines 1 and 3). Whereas, in the case of WT only a small difference between total and membrane protein was observed. Indeed, the ability of Qm to reach the membrane is significantly impaired (Fig. 4B). Similar results were obtained treating cells transfected with CD98-WT construct with tunicamycin before brefeldin assay. Indeed, tunicamycin treatment abolished in vivo CD98 N-glycosylation impairing the trafficking of the unglycosylated protein to the plasma membrane (Fig. S1).
Therefore, to further investigate the trafficking process, only the Qm was used in comparison with the WT. To follow the process of protein migration towards the membrane, HEK293 transiently transfected with WT or Qm construct, were treated with brefeldin that blocks the vesicle trafficking from Golgi towards the plasma www.nature.com/scientificreports/ membrane and used in confocal analyses 33 . In order to find optimal detection times after brefeldin treatment, the stability of the protein was firstly assessed by treating cells with cycloheximide, a well acknowledged inhibitor of protein synthesis. The amount of WT protein was still higher than 50% after 6 h of protein synthesis block ( Fig. 5A and Table S1); whereas, in the case of the Qm a strong decrease was detected already after 2 h and, after 6 h, the protein was less than 20% of the initial amount ( Fig. 5A and Table S1). This data demonstrated that N-glycosylation is important for protein stability. Moreover, data from Fig. 5A suggested to not prolong the observation in confocal analysis for more than 2 h after brefeldin treatment. As shown by Fig. 5B and Fig. S2, at time zero, most of the protein was in the cytosol both in the WT and Qm expressing HEK293 cells. After 1 h and, more evidently, 2 h, the WT mostly reached the plasma membrane, whereas the Qm was still localized into the cytosol, clearly demonstrating that in the absence of N-glycosyl residues, the protein capacity to migrate towards the plasma membrane was strongly impaired. The localization of WT and Qm before brefeldin treatment was shown in the upper panel of Fig. S2 as control.  www.nature.com/scientificreports/ Given that CD98 can form a functional heterodimer with LAT1, the effect of CD98 N-glycosylation on the trafficking of LAT1 towards the plasma membrane was also investigated. HEK293 cells were transiently co-transfected with LAT1 and alternatively with WT-CD98 or Qm-CD98; LAT1 was mainly localised in at the plasma membrane (Fig. S3A). After co-transfection, the treatment with brefeldin was performed to follow protein migration towards the plasma membrane. As shown by Fig. 6 and Fig. S3, LAT1 followed CD98 routing to the plasma membrane. In the case of co-transfection with WT-CD98, after 2 h from the brefeldin removal, LAT1 reached the plasma membrane. On the contrary, in the case of co-transfection with the Qm-CD98, LAT1 was mainly retained in the cytosol. To exclude any influence of the endogenous CD98 on the LAT1 trafficking, HEK293 cells were transfected with the sole LAT1 construct. In this case, the over-expressed LAT1 lost the capability to reach the plasma membrane (Fig. S4). Indeed, the endogenous expression of CD98 is negligible when compared to the overexpressed one ( Fig. S5 left panel). Moreover, the co-transfection of CD98 and LAT1 lead to the expression of both the protein in a good and comparable amount (Fig. S5 right panel).
These data were confirmed by biotinylation assay; indeed, only the membrane fraction derived from cells transfected with WT-CD98 also contained LAT1 (Fig. 7A).  www.nature.com/scientificreports/ In agreement with these findings, a transport assay was performed to test LAT1 function using one of the preferred LAT1 substrates, histidine 18 . The uptake of radiolabelled histidine was followed in cells transfected with WT-CD98 or Qm-CD98 (Fig. 7B). In good agreement with analyses performed on biotinylated samples (Figs. 6 and 7A), histidine uptake was significantly lower in cells transfected with Qm-CD98 compared to cells transfected with WT-CD98.

Discussion
In the present work, the role of N-glycosylation on the trafficking of CD98 towards the plasma membrane was assessed. The strategy for targeting CD98 made use of an anti-CD98 antibody against an epitope that is exclusively located on the extracellular side of the protein and an anti-FLAG antibody which, on the contrary, specifically binds to an intracellular site (Fig. 1). This allowed us to unequivocally visualize the protein which was located at the membrane surface (without permeabilizing cells), deriving from the endogenous protein plus the overexpressed one. On the other hand, the anti-FLAG staining (after cell permeabilization) highlighted the sole over-expressed protein, harbouring the FLAG tag at the N-terminus (Fig. 1). The virtual complete overlapping of the two staining protocols indicated that the endogenous protein does not interfere with the staining of the over-expressed one. Most importantly, the overlap of the two staining signals, confirmed that the over-expressed protein normally reaches the definitive location and that, the FLAG tag does not interfere with the overall trafficking mechanism; indeed, no accumulation of FLAG signal was unspecifically observed in intracellular organelles or the cytosol. This data validated both the over-expression system for the study of CD98 localization and the use of mutants for investigating the role of N-glycosylation, in the absence of interferences from the WT endogenous protein. Moreover, to better appreciate the role of N-glycosyl residues in the dynamics of CD98 trafficking, we employed a well-acknowledged combined approach of brefeldin and cycloheximide 33,34 that block cell trafficking and de novo protein synthesis, respectively. Therefore, from both WB analysis and confocal microscopy of WT and CD98 site-directed mutants some milestones were reached: (i) N-glycosylation was certainly involved in the protein trafficking to the cell surface (Figs. 3 and 4); (ii) most of the N-glycosylation moieties had to be removed to sensibly reduce the protein residence at the cell surface (Fig. 4), as demonstrated by the little effect of a single mutation on the overall trafficking process (Fig. 3); (iii) N-glycosyl moieties were also involved in the stabilization of the protein at the plasma membrane. Indeed, the lack of N-glycosylated residues triggered a much faster protein degradation, according to its improper localization (Fig. 5A). This finding is in good agreement with the hypothesis that the un-glycosylated protein is more accessible by proteases according to previous findings 35 . Thus, it can be hypothesized that the spots present in HEK293 cells transfected with the CD98-QM construct (Fig. 3B), result from a partial retention into the endoplasmic reticulum, and then followed by proteasomal degradation as previously suggested 30 . Another aspect is that, at least in the analyzed cell context, the major determinant of glycosylation process of CD98 is the N-glycosylation, since the PNGase treatment caused a dramatic shift of the apparent molecular mass of CD98 to that of the sole protein moiety (Fig. 2). We cannot exclude that in other cell contexts, both physiological and pathological, other glycosylation moieties, such as O-glycosylation, may concur to the Post-Translational Modifications (PTMs) of CD98. Bioinformatic predictions obtained through the NetOGlyc-4.0 server, which produces neural network predictions of mucin type GalNAc O-glycosylation sites in mammalian proteins, suggests the presence of O-glycosyl moieties attached to the S504. Furthermore, GlyGen, a database that retrieves information from multiple international data sources, shows The obtained samples were tested for the presence of native LAT1 using an anti-LAT1 antibody 18 . WT and Qm mutant were also tested to ensure that sufficient amount of membrane protein fraction was loaded in the gel using anti-FLAG antibody 33 . The western blot image is representative of three independent experiments. In (B), Uptake of [ 3 H]-histidine in HEK293 cells. HEK293 cells were seeded and transfected with WT-CD98 or Qm-CD98 constructs as described in materials and methods; transport was started by adding 5 μM [ 3 H]histidine to transfected cells. The transport was blocked after 1 min, by placing cells on ice and washing them with cold transport buffer as described in materials and methods. Results are means ± SD of three independent experiments. Significantly different from WT as estimated by Student's t-test. www.nature.com/scientificreports/ the presence of an O-linked glycan residue in CD98; however, this information comes from proteomic studies that didn't clarify the exact location of the O-Glycosylated residue. It has to be stressed that, O-glycosylation is mainly related to cell processes other than trafficking. This PTM was described in glycocalyx formation, hostmicrobes interaction, protection from proteolytic cleavage 36 and in several pathological conditions, including cancer 37 , and cardiovascular disease 38 .

Conclusion
Taken together our results, we can conclude that the major role of N-glycosylation in routing CD98 to the plasma membrane is here definitively demonstrated. The importance of understanding the molecular basis of CD98 trafficking lies for different reasons: as stated in the introduction, CD98 is a multifunctional protein and its presence/stability in the plasma membrane may affect several cell processes. In pathological conditions, such as cancer, CD98 has a pro-proliferative role due to interaction with integrin and adhesion pathways. Therefore, the possibility of preventing or blocking CD98 residence at the cell surface by acting on its trafficking, may result in slowing down uncontrolled cell growth/metastasis 39 . Moreover, a role in cell metabolism is also expected considering the formation of a functional heterodimer with some of the SLC7 members. These transporters are unglycosylated proteins and the chaperone role of CD98 in routing them to the plasma membranes was already proposed 19,27,40 . In this scenario, our data are in perfect agreement with the previous observations; more importantly, we here clarify the molecular basis of the chaperone function played by CD98. By combining confocal analysis, biotinylation and transport assays we demonstrated that one of the most studied SLC7 members, i.e., LAT1, follows CD98 to reach the plasma membrane. We may hypothesize that a similar scenario may occur in the case of the other SLC7 members known to interact with CD98, given that the heterodimer is formed at the level of two conserved residues 17,21 . This finding opens also other perspectives when considering that a therapy acting on CD98 trafficking will surely impair the LAT1 chaperoning, with consequent high relevance to human therapy. Then, CD98 inhibition would act synergistically with specific LAT1 inhibitors, such as the tyrosine derivative JPH203 that is in human trial for anticancer therapy 41 .

Materials and methods
Cloning and mutagenesis. The

SDS-polyacrylamide gel electrophoresis and western blotting. HEK 293 cells were solubilized
in RIPA buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% TX-100, 1% sodium deoxycholate) supplemented with protease inhibitors. The proteins were separated on a 12% polyacrylamide gel and transferred to a nitrocellulose membrane. After blocking, the membrane was incubated with a mouse anti-FLAG (1:1000) or a rabbit anti-LAT1 (1:1000) overnight at 4 °C, respectively. The blot was washed three times at room temperature and incubated with the appropriate secondary antibody (1:10,000) for 1 h and revealed by chemiluminescence.
In vitro endoglycosidase treatment. Cells expressing WT and mutants were solubilized as described above. Fifty μg of total lysate were treated with a denaturing buffer and incubated for 5 min at 90 °C. After cooling, N-glycosidase F from Merck was added to the lysate and left for 3 h at 37 °C. Aliquots were used for SDS-PAGE, followed by immunoblotting analysis.
Evaluation of protein stability. HEK293 cells were seeded on 6-well plates and cultured to 80% confluence. Then, cells were transfected with 1 µg WT and Qm constructs using PolyJet according to the manufacturer's procedures. Cells were incubated with cycloheximide (20 μg/ml) to inhibit protein synthesis. At increasing time points, cells were harvested, lysed, and subjected to SDS-PAGE followed by immunoblot analysis.
Confocal microscopy. HEK293 cells were seeded onto the coverslip and cultured until 60% of confluence.
According to the manufacturer's procedures, cells were transfected with 1 µg hLAT1, hCD98 WT and mutant constructs using PolyJet transfection reagent. After 24 h, cells onto the coverslip were fixed with 4% (w/v) paraformaldehyde (15 min; 37 °C), rinsed three times with PBS, and incubated with blocking buffer (10% bovine serum albumin in PBS) for 1 h at room temperature. Then, cells were incubated with the rabbit monoclonal anti-CD98 (1:250) as internal proof that the protein is located at the plasma membrane. The recognized epitope protrudes towards the extracellular milieu; hence, the antibody can also work without membrane permeabilization. After treatment with the first antibody, samples were extensively washed. Membrane was permeabilized with 0.1% Triton-X100, and incubation for 1 h with mouse monoclonal anti-FLAG antibody (1:500), whose epitope is located inside the cell (Fig. 1), was performed. The LAT1 labelling was performed with a post-permeabilization incubation of cells with mouse monoclonal anti-HA (1:500 Tunicamycin treatment. HEK293 cells were seeded on poly-D-lysine-coated 6-well plates and cultured up to 80% confluence. Then, cells were transfected with 1 µg hCD98 WT using PolyJet transfection reagent. After 6 h of transfection, tunicamycin (1 µg/mL) or vehicle (DMSO), was added to the complete medium and cells were grown up to 24 h. After tunicamycin incubation, cells were subjected to biotinylation procedure, as described below.
Biotinylation. HEK293 cells were seeded on poly-D-lysine-coated 6-well plates and cultured up to 80% confluence. Cells were transfected with 1 µg hCD98 WT and Qm constructs using PolyJet according to the manufacturer's procedures. After 24 h incubation, cells were washed twice in ice-cold PBS and incubated with 1 ml of 0.5 mg/ml biotin in PBS for 45 min at 4 °C. Then, cells were incubated for 30 min with PBS/Glycine buffer to quench all non-reacting biotin. After ice-cold PBS washing, cells were solubilized with RIPA buffer and incubated overnight with streptavidin agarose beads. Beads were then washed and centrifugated to isolate proteins were that located in the plasma membrane.
Other methods. The homology structural model of hCD98 was built through Modeller 10.2, using the Cry-oEM structure of CD98 21 , lacking the N-terminus, and the N-terminal ab initio model of CD98 from Rosetta. The glycosylation sites were predicted by NetNGlyc and NetOGlyc Server. The amount of protein in the total lysate was measured with the Lowry-Folin assay. The quantification of protein bands detected by immunoblotting was estimated by using the Chemidoc imaging system.

Transport measurements in cells. HEK 293 cells were seeded onto 12 well plates coated with Poly-D-
Lysine and cultured until they reached 60% confluence. Then, cells were transfected with Polyjet transfection reagent according to the manufacturer's procedures with 0.5 µg of WT-CD98 and Qm constructs. After 24 h transfection, cells were used for transport assay of [ 3 H]-histidine. In brief, cells were washed twice with a warm transport buffer, composed of 20 mM Tris HCl pH 7.5. Transport was started by adding radiolabelled 5 µM [ 3 H]-histidine and terminated after one minute by washing out the transport buffer and rinsing the cells three times with the same ice-cold transport buffer (0.5 ml per well per rinse). Then, cells were solubilized in 500 µL of 1% TX-100 solution, and 400 µL cell extracts were counted for radioactivity. The remaining 100 µL in each well was used for measuring total protein concentration. LAT1-histidine transport was evaluated by subtracting the transport values of each condition from those deriving from blank, i.e., samples treated with the well-known LAT1 specific inhibitor BCH. In particular, 10 mM BCH was added at time zero together with radiolabelled [ 3 H]-histidine in the transport buffer.

Data availability
The datasets analysed during the current study are available from the corresponding author on reasonable request.