N‐glycome inheritance from cells to extracellular vesicles in B16 melanomas

We investigated the correlation between metastatic behaviors of tumor cells and asparagine‐linked glycosylation (N‐glycosylation) of tumor‐derived extracellular vesicles (EVs). Three mouse melanoma B16 variants with distinct metastatic potentials show similar gene expression levels and enzymatic activities of glycosyltransferases involved in N‐glycosylation. All melanoma variants and EVs have nearly identical profiles of de‐sialylated N‐glycans. The major de‐sialylated N‐glycan structures of cells and EVs are core‐fucosylated, tetra‐antennary N‐glycans with β1,6‐N‐acetylglucosamine branches. A few N‐glycans are extended by N‐acetyllactosamine repeats. Sialylation of these N‐glycans may generate cell‐type‐specific N‐glycomes on EVs. Taken together, melanoma‐derived EVs show high expression of tumor‐associated N‐glycans, and the core structure profile is inherited during multiple selection cycles of B16 melanomas and from tumor cells to EVs.

the parent B16 cells (experimental metastasis) [12]. These variants show little spontaneous metastasis from subcutaneous tumors to the lungs [13]. Meanwhile, the B16-BL6 variant is selected for its high invasive ability following injection of the parent B16-F10 variant into the urinary bladder in vitro, resulting in a highly spontaneous metastatic variant [13].
Tumor cells secrete small vesicles, termed extracellular vesicles (EVs), that contain various cargo molecules including nucleic acids, soluble proteins, and membrane proteins [14]. Tumor-derived EVs promote tumor progression and metastasis by delivering their cargo molecules to surrounding tumor microenvironments and future metastatic organs [14]. Accumulating evidence has demonstrated that the molecular compositions of tumor-derived EVs are dependent on the metastatic potentials of their secreting tumor cells [15][16][17], suggesting a potential role of tumor-derived EVs as biomarkers for metastasis. Although aberrant Nglycosylation on tumor cells is known to promote tumor metastasis [6], little is known about the correlation between N-glycosylation of tumor-derived EVs and metastatic potentials of the secreting tumor cells.
In the present study, we investigated the correlation between metastatic potentials of three B16 variants (B16-F1, B16-F10, B16-BL6) and N-glycosylation of their EVs by characterizing N-glycan structures and gene expression, as well as enzymatic activity, of glycosyltransferases involved in N-glycosylation. The results demonstrate that the core structure profile of N-glycans is inherited at the genetic level during multiple in vivo and in vitro selection cycles of B16 variants and is copied from tumor cells to their EVs. It was suggested that sialylation of these N-glycans probably generates cell-type-specific N-glycome on EVs. This study establishes the N-glycosylation landscapes of B16 variants and their EVs, and indicates that the bulk N-glycosylation of melanoma EVs does not reflect the metastatic potentials of their secreting tumor cells in the B16 model.

Cell culture
Mouse B16-F1 and B16-F10 melanoma cells were purchased from the American Type Culture Collection (Manassas, VA, USA). B16-BL6 cells were purchased from RIKEN Bioresource Center. All cell lines were maintained in complete Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 100 UÁmL À1 penicillin, and 100 lgÁmL À1 streptomycin at 37°C in a 5% CO 2 atmosphere.

Preparation of EVs
Extracellular vesicles were prepared as described [18]. Briefly, melanoma cells (1 9 10 6 cells/10-cm dish) were cultured for 24 h, washed twice with PBS, and incubated for 48 h in EVdepleted medium. The conditioned medium was sequentially centrifuged at 130 g for 5 min, 20 000 g for 20 min, and 100 000 g for 70 min. The final pellet containing EVs was washed once with PBS and measured for its protein concentration with a BCA protein assay kit (Pierce, Rockford, IL, USA), using bovine serum albumin as an external standard.

Preparation of whole cell lysates
Cells were lysed for 15 min at 0°C in lysis buffer comprising 10 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, and protease inhibitor cocktail (Roche, Mannheim, Germany). The homogenate was centrifuged at 20 000 g for 10 min and the supernatant was recovered as the whole cell lysate.
Preparation and fluorescent labeling of N-glycans N-glycans were released from EVs and whole cell lysates by treatment with peptide-N 4 -(N-acetyl-b-glucosaminyl)asparagine amidase F (PNGase F) (New England Biolabs, Ipswich, MA, USA) according to the manufacturer's instructions. The released N-glycans were purified and fluorescently labeled with 2-aminopyridine (PA) using a BlotGlyco Kit (Sumitomo Bakelite, Tokyo, Japan) according to the manufacturer's instructions.

HPLC
PA-labeled N-glycans were initially separated into one neutral fraction and six sialidase-sensitive fractions by anionexchange HPLC, as described [19]. Sialidase sensitivity was assessed by incubating the anion-exchange HPLC fractions with 10 mU neuraminidase from Arthrobacter ureafaciens in 10 mM sodium acetate buffer (pH 5.5) for 16 h at 37°C. The sialidase-digested samples and the neutral fraction were desalted on PD-10 desalting columns (GE Healthcare, Chicago, IL, USA) and subjected to reversed-phase HPLC as described [19]. The major peaks (> 0.5% of total glycans) from the reversed-phase HPLC were further fractionated and analyzed by size-fractionation HPLC to determine the glycan structures and quantify their abundance using PA-labeled glucose hexamers in PA-glucose oligomers (degree of polymerization = 3-22; 2 pmolÁlL À1 ; Takara), as described [19].
Mass spectrometric analysis of PA-labeled N-glycans PA-labeled N-glycans fractionated by reversed-phase HPLC were subjected to matrix-assisted laser desorption/ionization  were mixed with 1 lL of 2,5-dihydroxybenzoic acid (10 mg/ mL in 50% acetonitrile/0.1% trifluoroacetic acid) on a target plate. After evaporation to dryness, spectra were obtained in the positive mode using an Autoflex mass spectrometer (Bruker Daltonics) operated in the reflector mode.

Structural determination of PA-labeled N-glycans from melanoma EVs
The structures of PA-labeled N-glycans from melanoma EVs were determined by reference to a database, glycoanalysis by the three axes of MS and chromatography (GALAXY) [20] version 2 (http://www.glycoanalysis.info/galaxy2/ENG/inde x.jsp), using the masses and glucose units of individual glycans, the latter of which were calculated by the elution positions of PA-glucose oligomers (degree of polymerization = 3-22) in the reversed-phase and size-fractionation chromatography [19].

Glycosyltransferase gene expression
Real-time PCR for glycosyltransferases was performed as described previously [21]. Briefly, total RNA was extracted from mouse melanoma cells (1 9 10 6 cells/10-cm dish) that had been cultivated for 48 h using TRI Reagent (Molecular Research Center, Inc., Cincinnati, OH, USA) in accordance with the manufacturer's protocol, followed by treatment with DNase (Qiagen) and purification using RNeasy Mini kit (Qiagen, Hilden, Germany). 1 lg total RNA was reverse-transcribed using an R 2 First Strand Kit (Qiagen) in a 40-lL reaction mixture and then diluted with 182 lL RNase-free water. The cDNA thus obtained was mixed with 2.7 mL RT2 SYBR Green qPCR Mastermix (Qiagen) and 2.496 mL water, and 25 lL of the mixture was applied to each well of a 96-well plate that contained specific primers for glycotransferase mRNAs (Qiagen). Two types of 96-well plates were used for quantification of 144 glycosyltransferase genes; a commercially available plate for Mouse Glycosylation (Cat. No. 330231 PAMM-046ZA) covering 84 glyco-related genes, and a 96-Well Custom PCR Array in which we manually selected 86 glycosyltransferase genes, seven house keeping genes (Actb, Gapdh, Hsp90a, Pgk1, B2m, Gusb, Ldha) and three primers for monitoring genomic contamination, cDNA synthesis and PCR reaction. cDNAs were amplified and analyzed using an ABI PRISM 7900HT thermocycler (Applied Biosystems, Waltham, MA, USA) according to the RT 2 Profiler PCR Array Handbook (Qiagen). The abundance of glycotransferase mRNAs relative to that of housekeeping genes (the average of Actb, B2m, Gapdh and Hsp90ab1) was calculated using the DC t method. The values in Tables 1 and Table S1 were calculated as the means of two independent experiments.

Glycosyltransferase activities in B16 variants
To correlate glycosyltransferase gene expression profiles and the enzymatic activities, we measured activities of Mgat3, Mgat4, Mgat5, Fut8, a2,3-and a2,6-sialyltransferases by using fluorescently labeled acceptor glycans (Fig. 1C). Consistent with the fact that Mgat3 gene was rarely expressed in all three B16 variants (Table S1), the enzymatic activity was found to be very low. Among B16 variants, B16-F1 cells showed the lowest Mgat3 activity, while this variant had higher Mgat4 and Mgat5 activities than B16-F10 and B16-BL6 cells. It is well known that Mgat5 is deeply involved in cancer metastasis [24] and that exogenous expression of Mgat3 inhibits experimental lung metastasis in murine model [26] because bisecting GlcNAc formed by Mgat3 has steric hindrance effects on the action of Mgat5 [27]. However, Mgat5 activity in B16 variants was very high, suggesting that basal levels of Mgat3 activity do not inhibit the action of Mgat5 in the cells. The Fut8 activity was high in B16-BL6 cells as compared with the other two variants. The a2,3-sialyltransferase activity was low in B16-F10 cells and similar between B16-F1 and B16-BL6 variants. The degree of enzymatic activity of Mgat3, Mgat4, Mgat5, Fut8 and a2,3-sialyltransferase (Fig. 1C) was in good agreement with their gene expression levels in B16 variants (Fig. 1B). However, a2,6-sialyltransferase activity in B16 variants was not associated with the gene expression pattern of St6gal1 and decreased over selection cycles. Together, these results indicate that B16 variants show unique glycosyltransferse profiles and imply that B16 variants have cell-typespecific N-glycome.

Identification of N-glycan structures of EVs from B16-F10 cells
To determine N-glycosylation profiles of EVs from B16 variants, they were prepared from the conditioned media by differential centrifugation. Consistent with our previous study on the extensive biochemical characterization of B16-F10-derived EVs (F10-EVs) [18], the EV preparation contained an EV marker protein CD81 ( Fig. 2A). F1-EVs and BL6-EVs were also prepared by the same procedures and found to express CD81 ( Fig. 2A). To compare the yield of EVs, we seeded the same number of cells in the same volume of culture medium, and obtained EVs from B16-F1 cells (1307 AE 465 lg protein), B16-F10 cells (1258 AE 140 lg protein) and B16-BL6 cells (578 AE 62 lg protein). Among the three B16 variants, the B16-F10 variant is positioned in the middle of the selection cycles [12,13]. Thus, we used this variant for extensive N-glycan analysis. N-glycans were released from F10-EVs by PNGase F digestion, fluorescently labeled, and separated into one neutral and six sialidase-sensitive fractions by anion-exchange HPLC (Fig. 2B). The number of sialic acids present on N-glycans of F10-EVs was similar to that of B16-F10 cells. These fractions were collected and analyzed by reversed-phase HPLC, revealing similar overall elution profiles between B16-F10 cells and F10-EVs (Fig. S2). However, some sialylated N-glycans in the Sia 2, Sia 3, Sia 4 and Sia 6 fractions were more enriched in F10-EVs than B16-F10 cells (Fig. S3), suggesting that sialylation probably generates EV-specific N-glycome.
Inheritance of core structure profiles of N-glycans during selection cycles of B16 variants and from the variants to their EVs To investigate whether characteristic core N-glycan structures were enriched in EVs, we compared reversed-phase HPLC profiles of de-sialylated complex-type glycans between F10-EVs and B16-F10 cells. For the comparison, we chose eight major peaks that contained nine N-glycans (Fig. 3A) and found that while the amounts of each N-glycan in F10-EVs were approximately 10fold larger than those in B16-F10 cells, the N-glycan profiles were very similar (Fig. 3A-D, F10-EVs and F10-Cells). These findings indicate that the core N-glycan structures and their relative abundance are inherited from B16-F10 cells to F10-EVs.
Next, we investigated whether the N-glycosylation profiles of EVs differed among the B16 variants. Anion-exchange HPLC analysis of sialylated glycans obtained from F1-EVs, F10-EVs, and BL6-EVs revealed that although the number of sialic acids  present on N-glycans was similar among the three variants, the Sia 3 fraction showed distinct elution patterns (Fig. 3E). Inconsistent HPLC profiles of the Sia 2 fraction between Figs 1C and 2E occurred for unknown reasons. The core structures of de-sialylated N-glycans showed no marked differences among the three EVs (Fig. 3A-D). We examined whether core structure profiles of Nglycans were inherited from B16-F1 and B16-BL6 variants to their EVs, and found that F1-EVs and BL6-EVs had nearly identical N-glycosylation profiles to their secreting tumor cells, except that slightly higher expression of tetra-antennary glycans in B16-F1 cells than the other variants was not inherited to the EVs (Fig. 3A-D). The high expression of tetra-antennary glycans in B16-F1 cells was predictable from higher expression of Mgat4 and Mgat5, and lower expression of Mgat3, the enzyme of which is known to inhibit the action of Mgat5 [26,30], than the other two variants (Fig. 1B,C). Together, these findings indicate that the core structure profiles of N-glycans are maintained during multiple in vivo and in vitro selection cycles of B16 variants, and mostly inherited from tumor cells to their EVs during EV generation. Our data also imply that sialylation probably generates cell-type-specific N-glycome on EVs.

Conclusion
The results of the present study provide the first detailed structural and quantitative comparisons of Nglycans expressed in EVs and their secreting tumor cells using three B16 variants with distinct metastatic potentials. Although B16 variants showed unique profiles of gene expression and enzymatic activity of glycosyltransferases involved in N-glycosylation, the overall core structure profiles of N-glycans were similar between the variants. Given the critical roles of sialylation in tumor progression, it is interesting to note that B16 variants expressed the same set of sialyltransferase genes involved in N-glycan modification. However, our data suggested that sialylation of N-glycans probably generates unique N-glycome in B16 variants and the EVs. Further extensive studies will be needed to elucidate the precise sialylation patterns of tumorderived EVs and their roles in tumor metastasis. It was shown that B16-F1 and B16-F10 cells both require N-glycosylation for adhesion to endothelial cells, as well as experimental lung metastasis [31,32]. Interestingly, the highly metastatic B16-BL6 variant had similar N-glycosylation profiles to the other variants, implying a general pathological role of this posttranslational modification in the establishment of lung metastasis. Experimental lung metastasis was reported to be promoted by EVs from highly metastatic melanoma cells and this EV function was clearly dependent on expression of hepatocyte growth factor receptor (Met) [15]. Although Met is an N-glycosylated protein, EVs from poorly and highly metastatic B16 variants shared the same core N-glycosylation pattern, indicating that the bulk N-glycosylation of melanoma-derived EVs does not reflect the metastatic potentials of their secreting tumor cells.
In conclusion, this study establishes the N-glycosylation landscapes of tumor-derived EVs and their secreting tumor cells in B16 models, enabling further exploration of the functions of the N-glycans on tumor-derived EVs in future studies.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Relative gene expression levels of 144 glycosyltransferases in B16 variants. Fig. S2. Comparative analysis of sialylated N-glycans from B16-F10 cells and F10-EVs. Fig. S3. Relative amounts of sialylated N-glycans in the Sia 1-6 fractions of B16-F10 cells and F10-EVs. Table S1. mRNA abundances of glycosyltransferases relative to the mean abundance of four housekeeping genes (Actb, B2m, Gapdh, and Hsp90ab1) in B16 variants. Table S2. Structural analysis of N-glycans expressed on EVs from B16-F10 cells.