Comprehensive Membrane N-Glycoproteomics Using Human Breast Cancer Cell Line Pairs

Aberrant glycosylation of membrane proteins is a hallmark of cancer and a useful molecular marker for the diagnosis of breast cancer (BC). However, the molecular mechanisms by which altered glycosylation affects the malignant transformations associated with BC are poorly understood. Accordingly, we performed comparative membrane N-glycoproteomics using the human BC cell line pair, Hs578T, and its syngeneic normal cell line, Hs578Bst. A total of 359 N-glycoforms derived from 113 proteins were identified in both cell lines, of which 27 were found only in Hs578T cells. Significant changes in N-glycosylation were found in the lysosome-associated membrane protein 1 (LAMP1), the integrin family, and laminin. Confocal immunofluorescence microscopy images revealed the accumulation of lysosomes in the perinuclear space in cancer cells, which could be associated with marked changes in LAMP1 glycosylation, such as a decreased level of polylactosamine chains. Overall, the alterations in glycosylation may be involved in changes in the adhesion and degradation of BC cells.


INTRODUCTION
e prevalence of breast cancer (BC) has increased over the last few years. In fact, BC is now one of the most common malignancies that a ect women worldwide. 1) BC is classi ed into heterogeneous subtypes, leading to diverse pathological manifestations associated with complex molecular mechanisms. 2) No single therapeutic approach is universally bene cial for BC owing to its diverse clinical course. As a result, various omics approaches, including genomics, 3,4) transcriptomics, 5) and proteomics, 6) have been utilized so as to better understand the molecular nature of BC.
Protein glycosylation is a major post-translational modication in mammalian cells. Nearly all plasma membrane proteins are glycosylated, 7) and their carbohydrate moieties play important roles in various biological events, such as adhesion, cell recognition, cell-cell interactions, and signal transduction. [8][9][10] Indeed, there is increasing evidence to show that glycosylation has an important role in cancer cell proliferation, invasion, and metastasis. 11,12) Aberrant core fucosylation of the cell surface receptor CD276 antigen is a profound immunosuppressive factor in BC. 13) High mannose and complex-type N-glycans, which increase BC progression, are involved in cell growth signaling, the ac-tivation of invasion and metastasis, and tumor-promoting in ammation. 14) e dysregulation of mannosyloligosaccharide 1,2-α-mannosidase, an oligosaccharide, is associated with the progression of BC. 15) However, the molecular mechanisms by which altered glycosylation a ects the malignant transformation of BC are poorly understood.
Glycoproteomics using mass spectrometry (MS) is the most useful method for analyzing glycosylation changes associated with the malignant transformation of cancer. In general, N-glycopeptides enriched by hydrophilic interaction chromatography, 16,17) lectin a nity, [18][19][20] and acetone 21,22) are subjected to liquid chromatography/tandem mass spectrometry (LC/MS/MS) to not only the deduce peptide sequence, but also the glycosylation site and glycan composition. A comprehensive analysis of the changes in glycosylation during cancer progression using MS-based glycoproteomics would be expected to clarify the role of glycans in the malignant transformation of cancer.
Cancer-derived cell pairs matched with normal cells from the same patient are valuable resources for cancer research. In BC, Hs578T and its syngeneic normal cell line, Hs578Bst, are the only cell pairs with the same tissue source of normal and cancer cells and have been well characterized. [23][24][25] Hs578T and Hs578Bst serve as novel and promising model systems for studying the molecular mechanisms of cancer. 26) In this study, we conducted comparative Nglycoproteomics to determine whether N-glycosylation contributes to malignancy using the human BC cell line pairs, Hs578T and Hs578Bst.

Trypsin digestion and glycopeptide enrichment
e acetone-based glycopeptide enrichment method was conducted as described previously. 21) Integral membraneassociated proteins were extracted from the two cell lines using the ProteoExtract Native Membrane Protein Extraction Kit (Calbiochem; Merck, Darmstadt, Germany). All membrane proteins were quanti ed using the bicinchoninic acid method, and 30 µg of the membrane proteins with 1% sodium deoxycholate were reduced with 10 mM dithiothreitol at 60°C for 40 min and alkylated with 20 mM iodoacetamide at 25°C in dark for 30 min. e membrane proteins were subsequently digested with 1.5 µg of modied trypsin (Promega, Madison, WI, USA) for 16 h at 37°C. Glycopeptides were precipitated with a ve-fold volume of ice-cold acetone, incubated for 16 h, and then centrifuged at 12,000×g for 10 min. A portion of the precipitate from both cell types was dissolved in 30 µL of 20 mM phosphate bu er (pH 7.2). To release the N-glycans, the sample was incubated with 2U PNGase F (Roche, Basel, Switzerland) containing 50% glycerol 16 h at 37°C. e deglycosylated peptides were desalted using a C-18 Spin Columns (Pierce; ermo Fisher Scienti c) and then dried with a SpeedVac concentrator.

LC/MS/MS and identi cation of the N-glycopeptides
Peptides and glycopeptides were separated on an EASY-nLC 1000 ( ermo Fisher Scienti c) using an L-column2 ODS trapping column (0.3×5 mm, 5 µm; Chemicals Evaluation and Research Institute, Tokyo, Japan) and a nano HPLC capillary column (75 µm×120 mm, 3 µm, C18; Nikkyo Technos, Tokyo, Japan). e mobile phase consisted of water containing 0.1% v/v formic acid (Pump A) and acetonitrile containing 0.1% v/v formic acid (Pump B). e sample injection volume was 5 µL. e peptides and glycopeptides were eluted at a ow rate of 0.3 µL/min with a linear gradient from 0 to 45% B over 115 min. Mass spectra were acquired on a Q Exactive mass spectrometer ( ermo Fisher Scienti c) equipped with a Nanospray Flex Ion Source ( ermo Fisher Scienti c) operated in the positive ion mode. e Xcalibur 4.0 workstation ( ermo Fisher Scienti c) was used for MS control and data acquisition. e spray voltage was set at 2.0 kV, the capillary temperature was maintained at 250°C, and the S-lens (stacked ring ion guide) RF (radio frequency) level was 50. Full mass spectra were acquired using an m/z range of 350-2000 for deglycosylated peptides or 700-2000 with a resolution of 70,000. Product ion mass spectra were acquired against the 10 most intense ions using a data-dependent acquisition method with a resolution of 17,500, normalized collision energy of 27%, and exclusion of 17s.
Raw data les were processed using the SEQUEST search engine ( ermo Fisher Scienti c) for peptides and the Byonic search engine ver. 3.11 (Protein Metrics, Cupertino, CA, USA) for glycopeptides, integrated as a node in Proteome Discoverer ver. 2.4 ( ermo Fisher Scienti c).
e UniProtKB database for humans (status/2021/07) and N-glycan database (307 N-glycans) were used for searching the database. e search conditions were as follows: trypsin, maximum number of missed cleavages, 2; static modi cation, carbamidomethylation (C); dynamic modi cations, Gln>PyroGlu (N-term Q), oxidation (M), deamidation (N), precursor mass tolerance of 10 ppm, fragment mass tolerance of 20 ppm, and maximum number of N-glycosylations per peptide, 2. A er the search, all identi ed peptides were ltered at a false discovery rate (FDR) threshold of 1% using a target/decoy search strategy. e identi ed peptides were evaluated for reliability using the percolator node. Only high-con dence N-glycoforms, glycopeptides with di erent glycan structures attached to the same peptide backbone, were processed using Microso Excel ver. 2209 (Microso , Redmond, WA, USA).

Glycoproteomics data analyses
Label-free quanti cation of the N-glycoforms was performed using a combination of a Minora Feature Detector, Feature Mapper, and Precursor Ions Quanti er nodes in Proteome Discoverer 2.4 ( ermo Fisher Scienti c). e analytical conditions were as follows: Peptide to Use, unique; Precursor Quanti cation, area; and normalization mode, none.
e Gene Ontology (GO) of the de-glycosylated proteins was determined using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) Bioinformatics Resources 6.8 (https://david.ncifcrf.gov/ home.jsp). Principal component analysis (PCA) was used to compare the glycoform peak areas between Hs578T (BC) and the normal breast Hs578Bst cell samples. e log2 fold change for N-glycoforms were plotted against the −log10 of the p-value. Statistical signi cance was declared at FDR LogWorth (−log10 p-value) of 1.3 (equivalent of a p-value of 0.05). e presence of signal peptides and transmembrane (TM) segments was predicted using SignalP (ver.5.0, https://services.healthtech.dtu.dk/service.php?SignalP-5.0) and TMHMM Server (ver. 2.0, https://services.healthtech. dtu.dk/service.php?TMHMM-2.0). Statistical analyses were performed using JMP Pro 15 so ware (SAS Institute, Cary, NC, USA). To predict the interaction pattern of aberrant N-glycoproteins in BC cells, the protein-protein interaction (PPI) network was analyzed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org). 27) Changes in site-speci c glycosylation in aberrant glycoproteins were examined via manual spectral analysis. e N-glycoforms identi ed using the Byonic search engine were supported by manual assignments. e typical oxonium ions were annotated on the spectra, including m/z at 204.08 (HexNAc), 274.09 and 292.10 (NeuAc), 325.11 (Hex-Hex), and 366.14 (HexNAc+ Hex). e peptide and glycan masses of the glycopeptides were deduced from the molecular masses of the peptide ion carrying a single N-acetylglucosamine, which is commonly found to be more intense. e remaining glycoforms were determined from the mass intervals between the glycoforms, and their peak areas were calculated from the extracted ion chromatogram (XIC) and summed across all charge states of the N-glycoforms. e site-speci c glycan pro les were compared using mean normalized peak area values. e relative peak area of each glycoform was normalized by the mean of the peak area value for all glycoforms derived from the same glycosylation site to enables comparative analysis of changes in glycan pro les of proteins with di erent expression levels. 28) Immunoblotting e integral membrane-associated proteins (1.5 µg per well) were solubilized in EzApply (ATTO, Tokyo, Japan) and denatured for 5 min at 95°C. Following separation by SDS-PAGE using 5-20% precast gel (ATTO, Tokyo, Japan) under a constant current of 20 mA for 80 min, the proteins were transferred onto a PVDF membrane using a Lightning Blotter (PerkinElmer, Waltham, MA, USA). e PVDF membrane were blocked with 5% ECL Blocking Agent (Cytiva, Marlborough, MA, USA) in PBS containing 0.1% Tween 20 (PBS-T) for 1 h. A er washing three times with PBS-T, the PVDF membrane was incubated overnight at 4°C with the primary antibodies against lysosome-associated membrane glycoprotein 1 (LAMP1) (1 : 1,000 dilutions). e membrane was washed three times with PBS-T, and incubated with an HPR conjugated anti-rabbit secondary antibody for 1 h. e PVDF membrane was washed three times with PBS-T, the target glycoproteins were detected using ECL Western Blotting Substrate (Pierce, ermo Fisher Scienti c), and visualized by LAS-3000 Imager (Fuji lm, Tokyo, Japan).

Immuno uorescence confocal microscopy
BC and control cells were cultured in glass-bottomed 12-well plates at 37°C in the presence of 5% CO 2 for 48 h. Cells on glass coverslips were washed three times with PBS and xed with methanol (1 mL). A er two rounds of washing with PBS, the cells were blocked with 5% albumin in PBS for 16 h at 4°C and then incubated with the primary antibodies in 5% albumin in PBS for 2 h at 37°C. ermo Fisher Scienti c). Finally, the cells were visualized using a TCS SP8 confocal microscope (Leica, Wetzlar, Germany).

Glycopeptide enrichment and N-glycoproteomics
e strategy used in our glycoproteomic study with the two syngeneic cell lines, Hs578Bst and Hs578T, is summarized in Fig. 1. Membrane-associated proteins, which are the major glycoproteins, were subjected to reductive alkylation followed by tryptic digestion. erea er, the glycopeptides were enriched via acetone precipitation. A database search for the conversion of Asn to Asp revealed 419 potential glycopeptides from the 230 proteins in both samples. When these glycoproteins were used for Gene Ontology (GO) annotations, approximately 40% (86 proteins) of the identi ed proteins were classi ed as "plasma membrane proteins," whereas the remaining proteins were categorized into extracellular exosome and/or organelle membranes, such as the ER, Golgi, and lysosomes (Fig. 2a). Figure 2b shows a qualitative comparison of the N-glycoforms identied using LC-MS/MS. LC-MS/MS revealed 249 and 217 Nglycoforms in the BC and control cell samples, respectively. A total of 359 membrane-related N-glycoforms from 113 proteins were identi ed. Among these, 142 N-glycoforms were found in BC cells, 110 were found in the control, and 107 were found in both samples. e MS raw data les were deposited in the ProteomeXchange Consortium via the jPOST partner repository (http://jpostdb.org) with the dataset identi ers PXD040130.

BC-related N-glycoforms
To con rm the ability of N-glycoproteins to distinguish between BC and normal cell lines, the PCA algorithm was implemented on the peak area values traced for the precursor ions of the N-glycoforms (Fig. 3a). Based on the scatter diagrams, BC and normal cells have separate data distributions. Accordingly, PCA revealed signi cant di erences between the N-glycoform spectra of cell samples. Volcano plots revealed a comparative distribution of the peak area values traced for the precursor ions of the N-glycoforms in both cell samples (Fig. 3b). N-glycoforms with statistically signi cant di erences (≥5-Di erence, FDR Logworth ≥1.3) are indicated by red circles. us, in the 59 N-glycoforms from the 31 glycoproteins, the N-glycan moieties were changed in a region-speci c manner. Some of these proteins may contain inconsistencies in N-glycosylation, such as lack of transmembrane helices or secretory signal sequences. Glycosylation validation using TMHMM and SignalP server further narrowed down these proteins, leading to 51 glycoforms derived from 27 glycoproteins for PPI network analysis (Table 1).

Site-speci c glycosylation changes in BC cells
N-Glycosylation changes occur mainly in proteins involved in limited intracellular pathways, such as the PI3K-Akt signaling pathway and phagosomes. To assess the site-speci c N-glycosylation changes, we identi ed three aberrant N-glycoproteins, ITGA5, LAMC1, and LAMP1, involved in these intracellular pathways. Figure 5a shows a comparison of the mean normalized peak area values of site-speci c glycan pro les between BC and control cell samples. Representative changes in N-glycosylation included the trimming of terminal mannose (N1107 and N1395 of LAMC), an increase in bi-antennary complextype glycans (N868 of ITGA5), a decrease in the polylac-tosamine chain (N84 of LAMP1), and the conversion of high-mannose type glycans to hybrid-type glycans (N261 of LAMP1). Glycosylation of these remarkable proteins was supported by the product ion spectra (Fig. 5b). e typical oxonium ions are annotated on spectra including m/z at 204.08 (HexNAc), 292.10 (NeuAc), 325.11 (Hex+ Hex), 366.14 (HexNAc+ Hex), and 657.23 (HexNAc+ Hex+ NeuAc).
e ions corresponding to a peptide ion carrying a single N-acetylglucosamine, and glycan-related ions that were correctly interpreted, have a high share of the total intensity in their spectra. e relative expression levels of the remarkable glycoproteins were analyzed by comparing the average peak areas ratios of the deamidated peptide ions from the two cell lines (Table 2). Based on the tumor-tonormal (T/N) peak area ratio of deamidated peptide ions, ITGA5, LAMC1 were upregulated in cancer cells with a T/N of over 2.0, whereas LAMP1 was downregulated in the cancer cells with a T/N ratio of less than 0.7 (Table 2). Using the LAMP1 as representative glycoprotein, the quantitative changes between the cell lines analyzed by LC/MS/MS were supported by western blot analysis (Supplementary Figure S1). LAMP1, which shows a remarkable change in N-glycosylation, was selected and subcellular localization analysis was performed.

Subcellular localization of LAMP1 in BC cells
LAMP1 is a major protein in the lysosomal membrane and is expressed in the plasma membranes of malignant cells. To determine the in uence of dynamic changes in glycosylation, we examined the subcellular localization of LAMP1 in BC and control cells (Fig. 6). Confocal immuno uorescence of cathepsin D, which is used as a lysosome marker, revealed that lysosomes were distributed throughout the cells in the control group, but had accumulated in the perinuclear space in cells in the BC group. Co-localization of LAMP1 and cathepsin D was observed in both cell lines.

DISCUSSION
In this study, we conducted N-glycoproteomics in an attempt to elucidate the relationship between glycosylation changes and malignant transformation of BC cells. Glycopeptides were enriched from the membrane protein fractions of representative BC cell lines and their syngeneic cell lines using an acetone-based glycopeptide enrichment method. All of the N-glycoforms that were identi ed by the Byonic search engine were compared between the BC and control cells. As shown in Fig. 3, PCA using the peak area values of the identi ed N-glycoforms revealed separate clusters in BC and control cells. e di erences in Nglycoforms between the two cell types suggest that cancerspeci c glycan alterations are due to the aberrant substrate selection of various glycosyltransferases in BC cells. In this study, the N-glycoforms of 27 proteins were found to be signi cantly altered in the BC cell line. In addition, these 27 abnormal N-glycoproteins were found to be associated with the PI3K-Akt pathway and phagosomes. ese intracellular pathways are not independent of BC. PI3K-Akt is a major signaling pathway that is involved in regulating cell proliferation, survival, and metabolism, but is o en the most activated oncogenic pathway in triple-negative breast cancer Proteins marked in red are associated with the PI3K-Akt signaling pathway, and those marked in blue are associated with the phagosome.
(TNBC). [29][30][31] erefore, clinical trials based on PI3K-Akt inhibition are currently underway. 32) Further, the epithelial-mesenchymal transition of TNBC cells via the PI3K/ Akt pathway involves the N-glycosylation of cell surface proteins. 33) In contrast, autophagy is maintained at high levels in TNBC cells, and the knockdown of relevant genes signi cantly suppresses cell proliferation, colony formation, migration/invasion, and induces apoptosis of TNBC cells. 34,35) Cell cultures under di erent conditions may a ect protein glycosylation. However, it should also be noted that the changes in glycosylation are found in glycoproteins, which seem to be primarily related to the PI3K-Akt pathway and phagosomes.
As shown in Fig. 5a, the N-glycan pro les of LAMP1 had the most remarkable change in BC cells. Tetra-antennary N-glycans with a polylactosamine structure were found on Asn84 in LAMP1: Hex 8 HexNAc 7 dHex 1 NeuAc 2-4 . Our data are consistent with ndings reported in previous studies that showed that LAMP1 is one of the most important 1,6-branched polylactosamine-carrier proteins in chronic myelogenous leukemia cells. 36) A decrease in the polylactosamine structure, which protects LAMP1 from the action of lysosomal hydrolases, in BC cells may result in the degradation of LAMP1 and the concomitant destabilization of lysosomal function. 37) As shown in Fig. 6, lysosomes containing LAMP1 predominantly accumulated in the cytoplasm around the nucleus. e spatial distribution of lysosomes is involved in cancer cell metastasis and drug resistance. 38,39) ese irregular lysosomal distributions may not be independent of the remarkable changes in LAMP1 glycosylation.
In conclusion, we employed N-glycoproteomics to demonstrate the aberrant N-glycosylation in BC cell samples and to identify the intracellular pathways a ected by this process. Membrane protein N-glycosylation was found to clearly di er between BC and control cell samples, and aberrant N-glycosylation at speci c sites may contribute to cell proliferation and survival. Further, the diversity of sitespeci c glycan changes suggests perturbation of glycosyltransferase substrate recognition in BC. Further validation using clinical tissue specimens and detailed site-speci c Nglycosylation analysis may enable the development of new therapeutics for BC.  Both cell types were simultaneously stained with cathepsin D (red) and LAMP1 (green) monoclonal antibodies. e nuclei were stained with 4′,6-diamidino-2-phenylindole (blue). Scale bars=17 µm.