Impact of Cystinosin Glycosylation on Protein Stability by Differential Dynamic Stable Isotope Labeling by Amino Acids in Cell Culture (SILAC)*

Cystinosis is a rare autosomal recessive lysosomal storage disorder characterized by intralysosomal accumulation of cystine. The causative gene for cystinosis is CTNS, which encodes the protein cystinosin, a lysosomal proton-driven cystine transporter. Over 100 mutations have been reported, leading to varying disease severity, often in correlation with residual cystinosin activity as a transporter and with maintenance of its protein-protein interactions. In this study, we focus on the ΔITILELP mutation, the only mutation reported that sometimes leads to severe forms, inconsistent with its residual transported activity. ΔITILELP is a deletion that eliminates a consensus site on N66, one of the protein's seven glycosylation sites. Our hypothesis was that the ΔITILELP mutant is less stable and undergoes faster degradation. Our dynamic stable isotope labeling by amino acids in cell culture (SILAC) study clearly showed that wild-type cystinosin is very stable, whereas ΔITILELP is degraded three times more rapidly. Additional lysosome inhibition experiments confirmed ΔITILELP instability and showed that the degradation was mainly lysosomal. We observed that in the lysosome, ΔITILELP is still capable of interacting with the V-ATPase complex and some members of the mTOR pathway, similar to the wild-type protein. Intriguingly, our interactomic and immunofluorescence studies showed that ΔITILELP is partially retained at the endoplasmic reticulum (ER). We proposed that the ΔITILELP mutation causes protein misfolding, ER retention and inability to be processed in the Golgi apparatus, and we demonstrated that ΔITILELP carries high-mannose glycans on all six of its remaining glycosylation sites. We found that the high turnover of ΔITILELP, because of its immature glycosylation state in combination with low transport activity, might be responsible for the phenotype observed in some patients.

Cystinosis is a rare autosomal recessive lysosomal storage disorder characterized by intralysosomal accumulation of cystine. The causative gene for cystinosis is CTNS, which encodes the protein cystinosin, a lysosomal proton-driven cystine transporter. Over 100 mutations have been reported, leading to varying disease severity, often in correlation with residual cystinosin activity as a transporter and with maintenance of its protein-protein interactions. In this study, we focus on the ⌬ITILELP mutation, the only mutation reported that sometimes leads to severe forms, inconsistent with its residual transported activity. ⌬ITILELP is a deletion that eliminates a consensus site on N66, one of the protein's seven glycosylation sites. Our hypothesis was that the ⌬ITILELP mutant is less stable and undergoes faster degradation. Our dynamic stable isotope labeling by amino acids in cell culture (SILAC) study clearly showed that wild-type cystinosin is very stable, whereas ⌬ITILELP is degraded three times more rapidly. Additional lysosome inhibition experiments confirmed ⌬ITILELP instability and showed that the degradation was mainly lysosomal. We observed that in the lysosome, ⌬ITILELP is still capable of interacting with the V-ATPase complex and some members of the mTOR pathway, similar to the wild-type protein. Intriguingly, our interactomic and immunofluorescence studies showed that ⌬ITILELP is partially retained at the endoplasmic reticulum (ER). We proposed that the ⌬ITILELP mutation causes protein misfolding, ER retention and inability to be processed in the Golgi apparatus, and we demonstrated that ⌬ITILELP carries high-mannose glycans on all six of its remaining glycosylation sites. We found that the high turnover of ⌬ITILELP, because of its immature glycosylation state in combination with low transport activity, might be responsible for the phenotype observed in some patients. Cystinosis (Online Mendelian Inheritance in Man 219800) is a rare autosomal recessive lysosomal storage disorder characterized by intralysosomal accumulation of cystine, leading to impaired function of multiple organs (1). Based on the age of onset and severity of symptoms, three clinical forms of cystinosis have been described: infantile (severe), juvenile and ocular (mild).
The causative gene for cystinosis is CTNS, which encodes the protein cystinosin, a lysosomal proton-driven cystine transporter (2,3,4). Recently, we have shown that in addition to transporting cystine, cystinosin is involved in nutrient-sensing in kidney proximal tubular cells via interactions with v-ATPase, Ragulator and Rags (5), and that the lack of cystinosin results in dysregulation of the mTOR signaling pathway.
Cystinosin is a 367-amino acid (AA) 1 protein with a nominal mass of 41.7 kDa. It has a short cytosolic 10-AA C-terminal tail, seven transmembrane (TM) domains spanning the lysosomal membrane and a 128-AA N-terminal region in the lysosomal lumen bearing seven N-glycosylation consensus sites N-X-S/T (N 36 , N 41 , N 51 , N 66 , N 84 , N 104 , and N 107 , hereafter referred to as N 1 to N 7 ) (Fig. 1A). The molecular weight (MW) of mature cystinosin is estimated at ϳ56 kDa. Cystinosin contains two lysosomal targeting motifs: a classical tyrosinebased motif (GYDQL) in the C-terminal tail and a conformational signal located on the fifth inter-TM loop (3,6).
Most early functional studies conducted on a set of 31 cystinotic mutations revealed a correlation between cystine transport activity and severity of symptoms (4). Most infantile mutations abolished cystine transport, whereas less severe mutations only reduced it (4). Patients affected by severe symptoms harbor severe mutations on both alleles, whereas patients with milder forms usually carry nonsevere mutations on both alleles or in association with a severe mutation.
However, some discrepancies have been observed, such as with I 67 -P 73 del (⌬ITILELP), which leads to 19% residual transport activity and is considered a nonsevere mutation (4). Individuals with the ⌬ITILELP mutation on both alleles have the juvenile form, whereas those with the ⌬ITILELP mutation on one allele in association with the 57-kb deletion (the most common severe mutation) can develop the infantile form (7)(8)(9). ⌬ITILELP is an in-frame deletion adjacent to an N-glycosylation site at position 66. The ⌬ITILELP mutation (deletion of AA67-73) is predicted to prevent the recognition consensus sequence of N-glycosylation N-X-S/T on NIT (AA66 -68) by oligosaccaryltransferase. Therefore, ⌬ITILELP results in the removal of seven amino acids in the N-terminal luminal region and the loss of glycosylation on N 66 , the fourth of seven N-glycosylation sites (10). To date, ⌬ITILELP is the only mutation detected in patients that leads to faulty glycosylation.
As glycosylation of lysosomal membrane proteins is known to protect from proteolysis (11), we hypothesized that faulty glycosylation caused by ⌬ITILELP might influence the stability of the protein. To investigate the stability of ⌬ITILELP, we compared its degradation rate to wild-type (WT) cystinosin. As additional controls, we also studied the degradation of two mutations that do not impact glycosylation sites (N 288 K and N 323 K). The classical approach for protein-turnover studies is pulse-chase labeling with radioactive amino acids. We used an alternative approach based on mass spectrometry, known as dynamic stable isotope-labeling by amino acids in cell culture (dynamic SILAC) (12,13). In this metabolic-labeling time-course experiment, the ratio of the unlabeled and labeled peptides of a protein reflects the ratio of pre-existing and newly synthesized proteins at each measured time point. It has been shown that from these values, the degradation rate of the proteins can be successfully calculated by curve-plotting or mathematical formulas (14,12).
To expand our knowledge of the cellular mechanisms involved in the pathogenicity of ⌬ITILELP, we analyzed its fate in the cell by studying its glycosylation, cellular local-ization and specific protein partners, as compared with WT cystinosin.

EXPERIMENTAL PROCEDURES
Antibodies-Antibodies were obtained from the following sources: antibody to PDI and calnexin from Enzo Life Sciences (Lyon, France), antibody to GFP from Roche, (Vélizy Villacoublay, France), antimouse IgG-HRP from GE Healthcare, and Alexa Fluor 555-and 647-conjugated secondary antibodies from Life Technologies (Thermo Fisher Scientific, Courtaboeuf, France). The Lamp-1 (1D4B) monoclonal antibody developed by J. Thomas August was obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the National Institute of Child Health and Human Development, and maintained by the Department of Biology at the University of Iowa, Iowa City, Iowa.
For dynamic SILAC studies, NIH/3T3 cells stably expressing different fusion proteins were plated at 4 ϫ 10 6 cells per 150-mm cell-culture dish. After removing the complete DMEM medium, the cells were quickly washed twice with PBS and then incubated with DMEM medium lacking arginine and lysine, supplemented with 175.2 mg/l [ 13 C 6 ] L-arginine, 84 mg/l [ 13 C 6 ] L-lysine, 10% dialyzed FBS (all from ThermoFisher), 100 mg/L proline (Sigma-Aldrich), 100 U/ml penicillin, 0.1 mg/ml streptomycin and 2 mM L-glutamine for chase periods of 0, 6, 18, and 24 h. At these times, the cells were treated for immunoprecipitation before being processed for proteomic turnover analysis.
To calculate the cell-growth rate (D), WT and mutant cells were plated in 10-cm tissue-culture dishes at 1.3 ϫ 10 6 cells per 150-mm cell-culture dish, cultivated for 24 h, collected at several time-points (0, 6, 18, and 24 h) and then counted using the Beckman Coulter Z2 cell counter. The cell growth rate (D) was calculated by fitting the number of cells (y) at different times (x) to an exponential growth curve in Microsoft Excel, using the formula: y ϭ a * eDx, where "a" is the number of cells at 0 h.
Generation of Stable Cell Lines-To generate NIH/3T3 fibroblast cell lines stably expressing human protein cystinosin-EGFP, the construct pCTNS-EGFP (3) was subcloned into the lentiviral pRRL. SIN.cPPT.PGK/WPRE vector (15). The cystinosin mutants ⌬ITILELP, N288K, N323K (5) and N1N7 (the seven putative N-glycosylated sites mutated in alanine) were obtained by the modification of the pRRL. SIN.cPPT.PGK/WPRE-CTNS-EGFP construct with specific mutagenic primers using the Stratagene QuikChange® Site-Directed Mutagenesis Kit according to the manufacturer's recommendations. Lentiviruses were produced in HEK293T cells as previously described (5). NIH/3T3 cells were transduced by lentiviral particles containing CTNS-EGFP or its mutated forms at an MOI of 7 in the presence of 8 g/ml polybrene.
Generation of EGFP Fusion Constructs and Transfection of NIH/ 3T3 Fibroblasts-The pCTNS-EGFP construct and its mutant form ⌬ITILELP have been described previously (3). For the mutant N66A (the N66 glycosylation site mutated in alanine), the modification of the pCTNS-EGFP construct (3) was carried out using the Stratagene QuikChange Site-Directed Mutagenesis Kit according to the manufacturer's recommendations. NIH/3T3 fibroblasts were transfected with 2 g of the different constructs using Lipofectamine® 2000 reagent (Invitrogen, Thermo Fisher Scientific, Courtaboeuf, France) according to the manufacturer's protocol. Cells were lysed 48 h after transfection for Western blotting.
Coimmunoprecipitation and Immunoblotting-Coimmunoprecipitation and Western blotting experiments on cystinosin-EGFP and its mutants in NIH/3T3 cells were performed as previously described (5). Immunoprecipitated calnexin was expressed as a ratio of the signals of calnexin and cystinosin-EGFP using BIO-1D software (Vilber Lourmat). Statistical analysis using the Kruskal-Wallis test (p Ͻ 0.05) with post-hoc Dunn's test was performed with Prism version 5 (GraphPad Software, La Jolla, CA). The threshold for statistical significance was set at p Ͻ 0.05. Each bar represents the mean Ϯ S.E. from four independent experiments.
Immunofluorescence-NIH/3T3 cells were plated on glass coverslips in 12-well tissue-culture dishes, then 24 -38 h later were treated as previously described (5). For the colocalization analysis, the reticular and lysosomal compartments were defined by delimiting the PDIor Lamp1-labeled areas, respectively, using ImageJ 1.50 software. The reticular or lysosomal accumulation of cystinosin-EGFP or its mutants was measured as a ratio, expressed in percentage of GFP fusion proteins that colocalized with the PDI-or Lamp1-labeled areas. Statistical analysis using the Kruskal-Wallis test with post-hoc Dunn's test was performed with Prism version 5 (GraphPad software). The threshold for statistical significance was set at p Ͻ 0.05. Three independent biological replicates were analyzed, and for each experiment, colocalization GFP/PDI or GFP/Lamp1 for five cells of each cell type was performed.
PNGase F and Endo H Treatments-Cell lysates were treated for 1 h at 37°C with 100 U of PNGase F or Endo H (BioLabs, Evry, France) per g of total proteins according to the manufacturer's recommendations except for the denaturation step, which was done at 50°C instead of 100°C. The resulting cell lysates were analyzed by Western blotting.
Drug Treatment and FACS Analysis-NIH/3T3 cells stably expressing WT cystinosin-EGFP and ⌬ITILELP-EGFP were incubated with 10 M clasto-lactacystin ␤-lactone (Calbiochem) or 100 nM bafilomycin A1 (SIGMA) for 24 h in the complete DMEM medium (Merck Millipore, Alsace, France). Cells were then harvested by using trypsin, and centrifuged for 2.5 min at 1500 rpm. The cell pellets were washed once with PBS and resuspended in 500 l PBS with 1% bovine serum albumin. EGFP expression was analyzed by direct flow cytometry using FACSCalibur with CellQuest software (BD Biosciences). Dead cells and debris were excluded by gating on forward/side light scatter, and 10,000 events corresponding to EGFP-positive cells were analyzed per sample. The average expression level of EGFP in individual NIH/3T3 cells was determined by calculating the geomean fluorescence intensity (GFI). The GFIs of the treated cells were normalized relative to the GFI of the untreated cells (with DMSO vehicle). Statistical analysis using the Kruskal-Wallis test with post-hoc Dunn's test was performed with Prism Version 5 (GraphPad software). The threshold for statistical significance was set at p Ͻ 0.05. Each bar represents the mean Ϯ S.E. from five independent experiments.
Mass Spectrometry (MS) Analysis for Turnover Study-For turnover studies, immunoprecipitated protein samples were resolved by SDS-PAGE on 10% gel and large protein bands corresponding to cystinosin were excised. In-gel trypsin digestion was performed as previously described (5). Nano-LC-MS/MS analysis of in-gel digested samples was performed on an Ultimate 3000 Rapid Separation Liquid Chromatography (RSLC) system coupled to a Q-Exactive Plus mass spectrometer (Thermo Scientific, Waltham, MA). Extracted peptides were resuspended in 0.1% (v/v) trifluoroacetic acid and 10% acetonitrile, then loaded onto a -precolumn (Acclaim PepMap 100 C18, cartridge, 300 m i.d.ϫ5 mm, 5 m, Dionex, Thermo Fisher Scientific, Courtaboeuf, France), followed by separation on the analytical 50-cm nano column (0.075 mm ID, Acclaim PepMap 100, C18, 2 m, Dionex). The chromatography solvents were (A) 0.1% formic acid in water, and (B) 80% acetonitrile with 0.08% formic acid. Peptides were eluted from the column using a gradient from 5% to 40% B over 38 min, then analyzed by data-dependent MS/MS, using the top-10 acquisition method. Briefly, the instrument settings were as follows: resolution was set to 70,000 for MS scans and 17,500 for the datadependent MS/MS scans to increase speed. The MS AGC target was set to 3.10 6 counts with a maximum injection time of 200 ms, whereas the MS/MS AGC target was set to 1.10 5 with a maximum injection time of 120 ms. Dynamic exclusion was set to 30 s.
Raw files were processed using the Proteome Discoverer 1.4.0.288 software (Thermo Scientific, San Jose, CA) and searched using Mascot 2.2 against the Mus musculus Uniprot KB/Swiss-Prot database v.5/7/2012 (16,331 entries) integrated with the sequences of the EGFP cystinosin WTs and mutants. Search parameters included fixed modification with carbamidomethyl (C) and variable modifications with oxidation (M), Label: 13 C(6) (K), Label: 13 C(6) (R), and two missed cleavages. The enzyme was trypsin, monoisotopic peptide mass tolerance was Ϯ2 ppm, and fragment mass tolerance was Ϯ0.05 Da. The identification validation was performed using a percolator, allowing 1-5% of target FDR for peptides based on the q-value and determined by the target-decoy approach.
The mass spectrometry proteomics data have been deposited at Panorama Public (https://panoramaweb.org/labkey/Guerrera.url). The MS raw files related to the turnover study have been deposited at the ProteomeXchange Consortium via the PRIDE (17) partner repository with the data set identifier PXD005357.
Turnover Data Analysis-The loss of a protein can be calculated from the relative isotope abundance (RIA) at each time point. The RIA represents the portion of the residual preexisting protein (MS1 area of the light peptides: A L ) versus the total amount of the protein in the cell, both pre-existing (A L ) and newly synthesized (MS1 area of the heavy peptides: A H ), at each time point. RIA is therefore calculated as: The MS1 areas A H and A L were extracted for peptides SVSLTVP-PVVK, EDGNILGHK, FEGDTLVNR and FSVSGEGEGDATYGK for cystinosin-EGFP using Skyline v2.6.0 (freely available at https:// skyline.gs.washington.edu). To ensure accuracy in the peak attribution, a dedicated Skyline library was built using the search files from the same samples. Peptide YFPQAYMNFYYK was excluded because of the presence of a methionine partially oxidized under different conditions. A thorough manual check and correction of peak picking was performed based on retention time and mass error. The RIA L,t was calculated for each peptide at each time point (0, 6, 18, and 24 h) for WT cystinosin and each mutants ⌬ITILELP, N 288 K, N 323 K, and N 1 N 7 (deglycosylated form of cystinosin) (12,16). The rate of loss of the protein, K loss , was calculated fitting the RIA L,t to an exponential decay curve using Excel and GraphPad v7: y ϭ e ϪKlossx K loss was additionally calculated mathematically in Excel, as the average of the K loss at each single time point for each peptide (14,16), in each independent biological triplicate: The rate of degradation (K deg ) was calculated as the difference between the rate of loss (K loss ) and the cell-growth rate (D): At least three independent biological replicates were analyzed (four for WT cystinosin-EGFP, three for ⌬ITILELP, N 288 K, N 323 K, and N 1 N 7 ). For each experiment, three to four peptides were measured at 0, 6, Table S1). Each mutant was always run in parallel with WT cystinosin-EGFP. The cell-growth rate (D) was measured in at least two independent experiments for each sample (four for WT cystinosin-EGFP, two each for ⌬ITILELP, N 323 K, N 288 K, and N 1 N 7 ) (details in supplemental Table S3). Statistical analysis using one-way ANOVA was performed with Prism version 7 (GraphPad software), using K loss values obtained by curve-fitting and formulas, from at least three independent experiments and three to four peptides per time point. The threshold for statistical significance was set at p Ͻ 0.05. Each bar represents the mean Ϯ S.D.

18, and 24 h (details in supplemental
Mass Spectrometry Analysis for Interactomic Study-For interactomic analysis, coimmunoprecipitated protein samples were concentrated on top of 10% SDS-PAGE gel and excised. In-gel trypsin digestion was performed as previously described (5). Analysis by LC-MS/MS was performed as described in "Mass spectrometry (MS) analysis for turnover study," except for gradient length, which was 120 min rather than 38 min. The mass spectrometry proteomics data and details on the identifications and quantifications have been deposited at the ProteomeXchange Consortium via the PRIDE (17) partner repository with the data set identifier PXD004948.
Interactomic data analysis of the MS files was processed with MaxQuant software version 1.5.2.8 and searched with the Andromeda search engine against the Mus musculus Uniprot KB/Swiss-Prot database 2016_01 (20199 entries), integrated with FASTA sequences of WT cystinosin-EGFP. To search parent mass and fragment ions, we set initial mass deviations of 4.5 ppm and 20 ppm, respectively. The minimum peptide length was set at seven amino acids and strict specificity for trypsin cleavage was required, allowing up to two missed cleavage sites. Carbamidomethylation (Cys) was set as a fixed modification, whereas oxidation (Met) and N-terminal acetylation were set as variable modifications. Match-between-runs was allowed using default parameters. The false discovery rate (FDR) for PSM and protein identification was set to 1%, determined by a target-decoy approach. Scores were calculated in MaxQuant as previously described, and left as the default parameters (18). The reverse-and common-contaminant hits were removed from the Max-Quant output. Proteins were quantified according to the MaxQuant label-free algorithm using Label-free Quantification LFQ intensities (19). Protein quantification was obtained using at least two peptides per protein. For pathway analysis, we annotated all of the proteins according to the KEGG database, and selected those classified as "protein processing in endoplasmic reticulum." Three biological replicates were analyzed for WT cystinosin-EGFP, ⌬ITILELP-EGFP and controls (without GFP). Statistical analysis between WT cystinosin and ⌬ITILELP was performed after the filteringout of all proteins identified the negative control with Perseus software (version 1.5.5.3, freely available at www.perseus-framework.org) (20). For statistical comparison, the LFQ data were transformed in log2 and imputed to fill in the missing data points by creating a Gaussian distribution of random numbers with a standard deviation of 33% relative to the standard deviation of the measured values, and a three-standard-deviation downshift of the mean to simulate the distribution of the low signal values. We performed a volcano plot analysis of the data using FDR ϭ 0.05 and S0 ϭ 0.5.
Experimental Design and Statistical Rationale-For the Western blot analyses, we performed four independent experiments. Statistical analysis was done with the Kruskal-Wallis test with post-hoc Dunn's test (threshold p Ͻ 0.05). For colocalization statistical analysis, we performed three independent biological replicates. Statistical analysis was done with Kruskal-Wallis and post-hoc Dunn's tests (threshold p Ͻ 0.05). For drug treatments and FACS analyses, we performed five independent biological replicates. Statistical analysis was done with Kruskal-Wallis and post-hoc Dunn's tests (threshold p Ͻ 0.05).
For MS turnover studies, we performed a minimum of three independent biological replicates (four for WT cystinosin-EGFP, three each for ⌬ITILELP, Asn 288 Lys, Asn 323 Lys, and Asn 1 Asn 7 ). For each experiment, three to four peptides were measured at 0, 6, 18, and 24 h (details in Table S1). Each mutant was always run in parallel with WT cystinosin-EGFP. The cell-growth rate (D) was measured in at least two independent experiments for each sample (four for WT cystinosin-EGFP, two each for ⌬ITILELP, N 323 K, N 288 K, and N 1 N 7 ) (details in supplemental Table S3). Statistical analysis using one-way ANOVA was performed with Prism version 7 (GraphPad software) using K loss values obtained by a formula (threshold p Ͻ 0.05) (details in supplemental Table S4A).
For MS-based interactomic studies, we performed three independent biological replicates for WT cystinosin-EGFP, ⌬ITILELP-EGFP and controls (cells without GFP expressing plasmid). Statistical analysis between WT cystinosin and ⌬ITILELP was performed with Perseus software (version 1.5.0.31). For statistical comparison, the LFQ data were transformed in log2 and imputed to fill in the missing data points by creating a Gaussian distribution of random numbers with a standard deviation of 33% relative to the standard deviation of the measured values, and with two to five standard deviations downshift of the mean to simulate the distribution of the low signal values. We performed a volcano plot analysis based on the t test, applying FDR ϭ 0.05 and S0 ϭ 1.

Detection of EGFP-tagged Cystinosin Peptides by MS-To
analyze the turnover of WT and mutated cystinosin, we transduced 3T3 cells with lentiviral constructs to stably express the cystinosin-EGFP fusion protein and its mutated forms. The schematic representation of WT cystinosin and ⌬ITILELP can be found in Fig. 1A and 1B. We consistently identified specific cystinosin peptides in accordance with the WT or the mutated forms N 288 K and N 323 K (supplemental Fig. S1). In all analyzed forms of cystinosin, we could identify the peptide AA23-33 (SVSLTVPPVVK). These data suggest that the signal peptide of cystinosin was cleaved after S22, generating a new N-terminal sequence starting with SVSLTVPPVVK, despite previous reports describing an uncleavable signal peptide (3,21). Our data agreed with the prediction of the currently used algorithm (http://www.cbs.dtu.dk/services/SignalP) (supplemental Fig. S2). ⌬ITILELP-EGFP mutation results in the loss of seven amino acids and one N-glycosylation site (N 66 ); consequently, its expected molecular mass would be ϳ3 kDa smaller than that of WT cystinosin, if all remaining sites were fully glycosylated. However, the observed molecular mass of the glycosylated form was much lower than expected and inconsistent with the simple deletion of seven amino acids and the loss of one glycan (Fig. 1C). Again, the detection of the peptide SVSLTVPPVVK (Fig. 1B) for ⌬ITILELP indicated that the N-terminal domain of the mutated protein was intact.
To further verify the effect of the glycan in position 66 (Asn 66 ) on the protein electrophoretic profile, we substituted the N-glycosylated site N 66 with alanine (N 66 A). N 66 A showed a molecular mass shift compared with WT cystinosin that was consistent with the loss of one glycan (Fig. 1C).
Half-life of Cystinosin and ⌬ITILELP-To investigate whether N-glycosylation influences the stability of ⌬ITILELP, we measured protein degradation using a dynamic SILAC approach. We incubated unlabeled cells expressing cystinosin and ⌬ITILELP with [ 13 C 6 ] L-arginine and [ 13 C 6 ] L-lysine for 6, 18, and 24 h in order to selectively label newly synthesized proteins. The WT cystinosin was used as the main control whereas N 288 K and N 323 K, which do not impact glycosylation, were used as additional controls. An artificial mutant simulating the loss of the seven glycosylations (N 1 N 7 ) was also investigated.
At each time point, cystinosin and its mutants were immunoprecipitated and analyzed by LC-MS/MS. For each sample, the MS analysis was performed to confirm the presence of cystinosin or its mutants and to verify the incorporation of labeled amino acids during the time-course. For calculating the turnover rate of a given protein, we used the precursor ion signal (MS1 area) from four labeled and unlabeled peptides from that protein at each time point (an example for one peptide is shown in Fig. 2). The extracted MS1 areas were used to calculate the relative isotope abundance at each time point (RIA t ) for each cell type and each replicate (supplemental Table S1). RIA L at time 0 was always set to 1 because the cells contained 100% of the light amino acids at this time point. The K loss values were obtained either by fitting (RIA L, t) on an exponential decay curve (K loss by formula) (12,13,16) or by single-time point analysis (K loss by fitting) (14, 16) (supplemental Table S2). The K loss obtained with the two ap-proaches fit a linear regression curve (R 2 ϭ 0.985), suggesting that the K loss values obtained by fitting on an exponential decay curve or on a single-time point analysis are very similar to each other (supplemental Fig. S3). The cells expressing cystinosin and its mutants were counted at each time point to determine the cell-growth rate or dilution rate (D) (supplemental Table S3).
The reproducibility of the measurements reported on the exponential decay curves between three different independent biological replicates appeared to be very high (Fig. 3) (16). The WT cystinosin was found to be very stable, with K loss ϭ 0.0434 Ϯ 0.008, whereas ⌬ITILELP was degraded three times faster than WT cystinosin, with K loss ϭ 0.1291 Ϯ 0.014 (Fig. 3, supplemental Table S2). The division rates of cells overexpressing cystinosin-EGFP and ⌬ITILELP-EGFP were similar, indicating that the loss rate of ⌬ITILELP was because of increased degradation (Fig. 3 histograms, supplemental Table  S4). N 288 K and N 323 K showed a tendency, though not significant, toward increased degradation when compared with WT cystinosin. Strikingly, the substitution of seven asparagine residues with alanine residues (N 1 N 7 ), mimicking complete deglycosylation, resulted in a very stable protein with a turnover similar to that of WT cystinosin (Fig. 3).
Cystinosin and ⌬ITILELP Lysosomal Degradation-To confirm the higher degradation rate of ⌬ITILELP, fibroblasts expressing ⌬ITILELP-EGFP or cystinosin-EGFP were treated for 24 h with bafilomycin A1 (Baf A1), an activity-inhibitor of lysosomes and other acidic compartments. EGFP fluorescence was then measured by flow cytometry. Upon Baf A1 treatment, the EGFP signal was significantly increased for both WT cystinosin and ⌬ITILELP when compared with DMSO-treated cells (vehicle), indicating that both proteins undergo lysosomal degradation. To test whether the proteasome partially contributes to the degradation of ⌬ITILELP, the same fibroblasts were treated with clasto-lactacystin ␤-lactone (CL), a specific proteasome inhibitor. In the fibroblasts expressing cystinosin-EGFP or ⌬ITILELP-EGFP, the EGFP fluorescence intensity did not change after 24 h of CL treatment, suggesting that proteasome degradation does not contribute significantly to the increased loss of ⌬ITILELP (Fig. 4).
ER Retention of ⌬ITILELP-To investigate whether the ⌬ITILELP mutation impacts protein-protein interactions, we explored the interactome of ⌬ITILELP compared with that of WT cystinosin by MS. We compared proteins coimmunoprecipitated by WT cystinosin or by ⌬ITILELP, then identified 181 such proteins as potentially specific interactants of WT cystinosin and/or ⌬ITILELP, as they were not identified in the negative control (without GFP) (supplemental Table S5). The results of this MS analysis largely confirmed our previously published interaction network generated for WT cystinosin (5), as we identified eight subunits of the vacuolar type H ϩ -ATPase (V-ATPase) and Rag GTPase C (Fig. 5, full protein list in supplemental Table S5). ⌬ITILELP maintained the same protein-protein interactions as WT cystinosin in the lysosome. However, ⌬ITILELP, but not WT cystinosin, specifically coimmunoprecipitated 17 proteins directly implicated in proteinprocessing in the ER (Fig. 5).
We investigated the cellular distribution of WT cystinosin and its mutants in cells stably overexpressing EGFP-tagged proteins by immunofluorescence. WT cystinosin-EGFP presented a predominant colocalization with the late endosomallysosomal marker Lamp-1 (83.62 Ϯ 1.73%) and a minor colocalization with protein disulfide isomerase (PDI), an ER marker (15.64 Ϯ 2.28%). Similar to WT cystinosin, N 288 K and N 323 K showed predominant lysosomal localization in 3T3 cells. In striking contrast, the signal for ⌬ITILELP corresponding to an ER localization was higher (48.26 Ϯ 2.57%), along with a decreased punctuated late endosomal-lysosomal localization pattern (47.66 Ϯ 2.01%) (Fig. 6A and 6B). Multitransmembrane proteins that do not acquire proper configurations accumulate in the ER, where they are recognized by calnexin (22). To analyze the interaction between cystinosin and this chaperone, lysates of 3T3 cells stably expressing cystinosin-EGFP or its mutated forms were immunoprecipitated with anti-GFP antibody, and coimmunoprecipitated endogenous calnexin was detected by Western blotting. As expected from the localization data, interactions with calnexin were significantly increased for ⌬ITILELP compared with WT cystinosin, N 288 K, or N 323 K (Fig. 6C and 6D). The mutants N 66 A and N 1 N 7 appeared to localize at the lysosome as well, without any ER retention (supplemental Fig. S4). Together, these data indicate that the ⌬ITILELP mutation leads to partial retention of cystinosin in the ER because of misfolding of the protein.
N-linked High-mannose Glycans of ⌬ITILELP-Lysosomal transmembrane proteins are synthesized in the ER, where folding and the initial stages of glycosylation (high-mannose type glycans) occur. They are then targeted to lysosomes via FIG. 3. Degradation curves of cystinosin and its mutants. RIA (relative isotope abundance, y axis) for each time point (x axis) were plotted using GraphPad software. The curve in black represents the degradation of WT cystinosin, whereas the curve in red represents the degradation curve of the indicated mutant. The histograms represent the loss rate (K loss ) for each protein and its two components: rate of degradation (K deg ) and cell-division rate (D). the trans-Golgi network, acquiring further modifications (complex type N-glycans). To verify the glycosylation stage of WT cystinosin and ⌬ITILELP, we performed enzymatic deglycosylation using the peptide:N-glycosidase F enzyme (PNGase F), which removes all N-linked glycan residues. As a control, we used N 1 N 7 , the constitutively deglycosylated form of cystinosin. After incubation with PNGase F, WT cystinosin had the migration profile of N 1 N 7 , whereas N 1 N 7 did not show any changes (Fig. 7A), showing that WT cystinosin is indeed heavily glycosylated. ⌬ITILELP was also completely deglycosylated by PNGase F, showing that it also contains a certain level of glycosylation. N 288 K and N 323 K, used as positive controls, showed the same electrophoretic profile as WT cystinosin.
We therefore tested the effect of the Endoglycosidase H enzyme (Endo H), as it specifically hydrolyzes nonmature N-glycans (high-mannose type). In contrast to WT cystinosin, ⌬ITILELP was sensitive to Endo H digestion (Fig. 7B), suggesting that ⌬ITILELP carries high-mannose on its available glycosylation sites. WT cystinosin is not affected by Endo H digestion, confirming that the WT form contains complex-type glycans, acquired during maturation in the Golgi apparatus.

DISCUSSION
In this study, we investigated the impact of cystinosin deletion ⌬ITILELP on the stability of the protein. The ⌬ITILELP mutation disrupts the proper glycosylation of cystinosin on N66. In the homozygous state, ⌬ITILELP leads to juvenile cystinosis, which correlates with the ability to maintain minimal cystine transport activity in vitro (9). However, when found in the heterozygous state together with a 57-kb de- letion on the second allele, the ⌬ITILELP mutation can result in the development of the severe infantile phenotype (7,8). These surprising observations led us to hypothesize that the deletion of seven amino acids, together with the lack of glycosylation on N 66 , interferes with cystinosin stability.
To investigate ⌬ITILELP turnover, we used the dynamic SILAC approach with MS. However, we faced two main challenges. First, cystinosin is not easily detectable by MS in nonfractionated samples, such as a total cell lysate or blood (no records on peptide identification in proteomic studies are reported at nextprot.org). This is probably because of the very low abundance of endogenous cystinosin in cells and to characteristics of its sequence, which contains seven TM domains and seven glycosylation sites. To our knowledge, endogenous cystinosin was identified with two peptides in only one MS study of enriched lysosomal membranes (23). Furthermore, no good antibodies are available to immunoprecipitate endogenous cystinosin from cell lysates in order to enrich endogenous cystinosin. For these reasons, we developed cellular models of stably expressing EGFP-tagged WT and mutated cystinosin. Although these tagged proteins localize correctly in the lysosome and have been previously used for functional studies, we are aware that the use of EGFP-tagged cystinosin is a constraint and we cannot be certain that the endogenous proteins follow exactly the same behavior. In this study, we detected four peptides of cystinosin (two from WT cystinosin and two specific to mutants) and provided evidence that the N-terminal signal peptide can be cleaved before S 23 .
The second challenge was that dynamic SILAC, although a well-established approach, can be tricky when studying the turnover of stable proteins. We calculated the K loss from the MS1 areas of the peptides and their labeled counterparts using curve-fitting or mathematical formulas applied to each time point (reported in different forms in the literature, but equivalent) (14,16). Our data confirmed that the K loss values calculated by curve-fitting and mathematical formulas correlate perfectly when many data points are used in the calculations. However, stable proteins such as WT cystinosin are virtually undegraded during the cell cycle/lifespan (rate of degradation: K deg ), and are only diluted upon cell division. In this case, the rate of loss (K loss ) of the protein is very close to the cell-division rate (D), and the K deg ϭ K loss -D value nearly reaches zero. After verifying that D was similar for all of the cell lines studied, we chose to report and compare the K loss of WT cystinosin and the mutants.
Our dynamic SILAC data showed that WT cystinosin is a very stable protein and that the ⌬ITILELP mutation increased 3-fold the degradation of the protein, whereas the N 288 K and N 323 K mutations did not significantly influence the protein's stability. We clearly showed that ⌬ITILELP undergoes lysosomal degradation, as v-ATPase inhibition by bafilomycin A1 (BafA1), but not proteasomal inhibition, resulted in increased cellular levels as observed by an increased fluorescence signal. The rate of lysosomal degradation of ⌬ITILELP was three times greater than that of the WT protein, further confirming the instability of this mutant.
We have recently demonstrated that cystinosin mutations, in addition to inefficient efflux of cystine from the lysosome, induce changes in the protein network that might modulate the role of cystinosin in the mTOR pathway (5). For this reason, we decided to investigate the ⌬ITILELP protein network. Our interactomic study revealed that similar to WT cystinosin, ⌬ITILELP maintains interactions at the lysosomal membrane with the V-ATPase FIG. 5. Protein partners for WT cystinosin and ⌬ITILELP in lysosomal and reticular compartments. Volcano plot representation of proteins immunoprecipitated by the anti-GFP antibody and identified by mass spectrometry in ⌬ITILELP-EGFPversus cystinosin-EGFPexpressing cells. In blue, proteins known to be lysosomal partners of cystinosin. In red, proteins classified by the KEGG pathway database as "protein processing in endoplasmic reticulum". complex and some proteins of the mTOR pathway. The presence of functional ⌬ITILELP on the lysosomal membrane aligns with the fact that patients with homozygous ⌬ITILELP mutations develop juvenile rather than infantile cystinosis. In addition, ⌬ITILELP interacts with 15 proteins of the ER, most of which have been implicated in ER quality-control and processing (e.g. calreticulin, Os9, Sel1l) (24). Immunofluorescence studies confirmed these findings and showed the partial localization of ⌬ITILELP in the ER and in lysosomes.
The accumulation of misfolded proteins in the ER can lead to ER stress and cell death. To prevent cell injury, nonnative proteins can exit the ER to undergo proteasomal and/or lysosomal degradation (25,26). To better understand the reasons for ER retention, we further explored the extent of the faulty glycosylation of ⌬ITILELP and demonstrated that it carried immature high-mannose glycans in all six remaining sites not directly affected by the mutation, suggesting that it is prevented from reaching the Golgi network for further processing. This is coherent with the observed ⌬ITILELP molecular weight by SDS-PAGE. Several studies have been conducted to decipher the glyco-code of processed proteins, revealing the importance of N-glycans in directing protein maturation (27). The glycan composition controls the calnexin binding cycle, which is important for the recognition of misfolded proteins in the ER (28,29). Along these lines, the strong interaction shown in our study between ⌬ITILELP and the ER chaperone, calnexin, suggested that this mutant was recognized as a nonnative protein by the ER quality-control system and was partially retained in this compartment. The N1N7 mutant, which does not carry any glycosylation, is correctly directed from the ER to the lysosome and its stability is not affected.
We suggest that ⌬ITILELP is initially retained in the ER, preventing it from being processed in the Golgi to obtain mature glycosylation, but it is directly transported to lysosome via its two targeting sequences. Further studies will be FIG. 6. ER retention of ⌬ITILELP. A, NIH/3T3 cell lines stably expressing cystinosin-EGFP, ⌬ITILELP, N288K and N323K were colabeled for PDI or Lamp-1 (confocal microscopy; scale bars ϭ 10 m). These images are representative of three biological replicates. B, Targeting of cystinosin-EGFP (WT and mutants), quantified as the percentage of the protein that is positive for PDI or Lamp-1 within five cells per group. The percentage of colocalization between cystinosin-EGFP and PDI or Lamp-1 was analyzed with the ImageJ 1.50 software (**, p Ͻ 0.005; *** p Ͻ 0.0001 as compared with WT cystinosin). C, NIH/3T3 cell lines stably expressing cystinosin-EGFP, ⌬ITILELP, Asn 288 Lys and Asn 323 Lys were immunoprecipitated with anti-GFP antibody, and coimmunoprecipitated calnexin was analyzed by western-blotting. Representative blots from four independent experiments are shown. D, Quantification of the calnexin signal normalized to the cystinosin signal (WT and mutants) (*, p Ͻ 0.05 as compared with WT cystinosin; each bar represents mean Ϯ S.E. from four independent experiments). necessary to fully understand the escape mechanisms of ⌬ITILELP from the ER to the lysosome. One possible mechanism is ER macro-autophagy that may be induced by ER stress (30). However, some unconventional cellular mechanisms of ER escape by retained proteins have also been described. Nonmature CFTR can be targeted to the plasma membrane, skipping the Golgi compartment via the unconventional GRASP-dependent pathway (31). Additionally, the formation of ER-derived vesicles referred to as EDEMosomes, which are necessary for ERAD tuning, enable the removal of selected ERAD regulators from the ER and their degradation by proteasomes and endolysosomal proteases. These vesicles contain EDEM1 and Sel1L but can also carry other ER proteins (32).
We conclude that the high turnover of ⌬ITILELP, because of its immature glycosylation state together with low transport activity, might be responsible for the phenotype observed in some patients who carry this mutation heterozygously, together with the 57-kb deletion.
Data are available via ProteomeXchange with identifier PXD004948, PXD005357 and on Panorama Public at https://panoramaweb.org/labkey/Guerrera.url. FIG. 7. Deglycosylation of cystinosin and its mutated forms. A, Lysates of NIH/3T3 cell lines stably expressing cystinosin-EGFP, N1N7 (seven N-glycosylation site-deleted mutants), ⌬ITILELP, N 288 K, and N 323 K were incubated with or without PNGase F for 1 h at 37°C to remove N-glycans. B, Lysates of NIH/3T3 cell lines stably expressing cystinosin-EGFP and ⌬ITILELP were incubated with or without Endo H for 1 h at 37°C to cleave specifically immature N-glycans. All lysates were resolved by SDS-PAGE on 10% gel and analyzed by immunoblotting with anti-GFP antibody.