1. Introduction
The growth of primary or metastatic tumors often leads to hypoxic conditions in their central regions, triggering cellular adaptation mechanisms [
1]. In triple-negative breast cancer (TNBC), these adaptive responses result in a more aggressive phenotype and increased resistance to chemotherapy and radiotherapy [
2]. Hence, understanding the tumor cell responses to hypoxia is crucial for developing more effective therapies to counter tumor progression and treatment resistance.
Extracellular vesicles (EVs) are nano-sized particles secreted by all cell types containing bioactive molecules such as proteins, DNA, and mRNAs, which play a crucial role in intercellular communication [
3]. Tumor-derived EVs, influenced by their microenvironmental conditions, have potential as therapeutic targets or biomarkers for specific cancer types and stages [
3]. Hypoxia-induced EVs carry cargo that can enhance pro-angiogenic signaling, promote epithelial-to-mesenchymal transition, facilitate metastasis, and may induce invasive behavior in breast cancer cells [
4].
Glycans, carbohydrate structures that modify proteins and lipids, are formed through a tightly controlled biosynthetic pathway involving glycosyltransferases and sugar transporters [
5]. These structures are predominantly found on the surface of cell membranes (glycocalyx) and are essential for cell–cell communication and interactions with the extracellular matrix [
6]. In cancer, significant changes occur in different types of glycosylation, notably N-glycans, which attach to asparagine residues within the sequon motif (Asn-X-Ser/Thr) and can be categorized into truncation, high-mannose, complex/hybrid, and other structural features [
7]. The surface proteins of EVs are highly glycosylated [
8,
9], and alterations in the N-glycan structures associated with tumor cells have been detected in cancer-derived EVs, indicating their potential utility as significant biomarkers for clinical applications [
6]. EVs expressing low levels of bisecting GlcNAc-modified N-glycan significantly contributed to the carcinogenesis and metastasis of the recipient cells, underlining the crucial impact of specific N-glycan structures on cancer progression [
10]. Hypoxia significantly alters the glycosylation profiles of breast cancer cells, leading to increased epithelial-to-mesenchymal transition and migration [
11]. However, the implications of N-glycosylation regarding the function of hypoxia-induced EVs remain unclear and necessitate further research.
The objective of this study is to investigate the impact of hypoxia on the N-glycan profiles from the secreted EVs of breast cancer cells in vitro and determine the molecular alterations underlying these changed N-glycosylation profiles. We lowered the oxygen concentration applied to the breast cancer cells to 1% (hypoxia) and isolated the EVs of an MDA-MB-231 cell line; the control cells were kept under atmospheric air (normoxia). Using liquid chromatography–tandem mass spectrometry (LC–MS/MS), we conducted quantitative differential proteomic and N-glycomic analyses to profile the EVs and their corresponding whole cell lysates. This comprehensive approach aims to elucidate the relationship between hypoxic cancer cells and N-glycan modifications, thereby identifying the potential molecular markers and therapeutic targets associated with breast cancer progression.
3. Discussion
Glycosylation is a complex, highly regulated post-translational modification that is associated with both physiological and pathological conditions [
14]. In highly proliferative solid tumors, low oxygen levels prompt cancer cell adaptation towards a more aggressive, chemoresistant, and radioresistant phenotype [
2]. Hypoxia significantly influences glycosylation enzymes, resulting in altered glycan profiles [
1]. Tumor-derived EVs enhance cancer cell adaptation to hypoxic conditions [
4], but the effects of hypoxic stress on N-glycan dynamics and their loading into EVs to modulate cancer progression are not well understood. In this study, we isolated EVs from breast cancer cells via ultracentrifugation and analyzed the role of N-glycosylation in the hypoxia-induced EVs using proteomics and N-glycomics via nano-LC–MS/MS. We characterized the N-glycan profiles of the EVs and cells under both normoxic and hypoxic conditions. Our results indicate that tumor hypoxia induces dynamic changes in the N-glycosylation-related proteins and cellular N-glycan profiles, which may affect tumor malignancy and potentially impact patient outcomes. Oxygen deprivation triggers the synthesis of specific subsets of EV N-glycans that profoundly alter cancer cell migration and invasive potential. These insights into hypoxia-induced N-glycosylation alterations identify potential biomarkers and therapeutic targets for cancer treatment.
Since glycosylation is non-template-driven, it is influenced by various factors, including the expression efficiency of glycosyltransferase enzymes within the ER or Golgi, sugar nucleotide transporters, and the availability of sugar nucleotides [
14]. Hypoxia affects the expression and functional activity of glycosyltransferases and the synthesis and transport of monosaccharides to the ER and Golgi, indirectly altering N-glycosylation [
1]. Under hypoxic conditions, 43 asparagine N-linked glycosylation-related proteins in a cell influence various aspects of N-glycosylation, such as biosynthesis, ER trimming, and the calnexin/calreticulin cycle, as well as ER to Golgi transport and sialic acid synthesis. Hypoxia also stimulates glycolysis in cells, subsequently altering the availability of crucial substrates for N-glycan synthesis, thus impacting key cellular functions and interactions [
1]. The dysregulation of the metabolic pathways is further exemplified by changes in critical metabolites like glucose, fructose, glucose-1-phosphate, and glucose-6-phosphate, which enhance glycolytic flux. Concurrent alterations in glycolysis/gluconeogenesis-related proteins in cells such as GAPDH, PGM1, and ALDOA influence substrate availability and nucleotide sugar synthesis, further affecting N-glycosylation and potentially leading to modifications in cellular behavior that are crucial for cancer progression.
Further understanding the effects of hypoxia on glycosylation and the resulting alterations in glycosylation-related proteins can provide critical insights into the N-glycan profiles produced by tumor cells, holding significant diagnostic and therapeutic potential. The analysis of the N-glycan patterns in breast cancer cells under hypoxic conditions demonstrates significant alterations closely tied to the mechanisms of cancer progression and metastasis. Utilizing LC–MS/MS, we distinguished 231 N-glycan patterns in the cells. N-glycan biosynthesis results in various glycosylation traits, such as branching, sialylation, and fucosylation, which critically influence cancer development by modulating cell signaling, adhesion, and the immune evasion mechanisms [
5]. Notably, the relative abundance of high-mannose glycans, specifically Man3 (H3N2F0S0) and Man9 (H9N2F0S0), was elevated in the hypoxic cells. These N-glycans are frequently associated with breast cancer metastases, with Man9 linked to worse clinical outcomes in high-grade tumors [
15,
16] and Man3 showing increased levels in benign breast tumor tissues compared to para-carcinoma tissues [
17]. Our results also highlight significant changes in the sialylation and fucosylation of N-glycans, emphasizing the importance of terminal modifications under hypoxia. For instance, the level of H3N5F1S0 was notably reduced in the hypoxic cells compared to in the normoxic conditions, a change implicated in the processes leading to breast cancer metastases to the brain [
18]. Conversely, H5N4F0S1 was elevated in those cells under hypoxic conditions, exhibiting patterns similar to those observed in highly metastatic cells [
18]. To investigate the underlying mechanisms, we analyzed the role of the proteins involved in the N-glycosylation-related pathways in cellular N-glycan alterations. These proteins showed high correlations with the dysregulated N-glycans in the hypoxic cells. Collectively, these findings illustrate that hypoxia significantly affects the expression of various N-glycan species by modulating the pathways involved in glycosylation biosynthesis. These N-glycosylation changes are potentially pivotal in facilitating brain metastasis and enhancing the overall metastatic progression in cancer.
EVs carry cargo that closely reflects a similar composition to their parent cells but can be modulated by microenvironmental factors such as hypoxia [
19]. The weak correlation reflected by a low Pearson’s correlation coefficient between significantly differential abundant cellular and vesicular N-glycans under hypoxia suggests that, during EV formation, selective sorting mechanisms are at play, allowing specific glycosylation patterns to be preferentially packaged into EVs, leading to different N-glycan patterns in cells versus EVs. This selective sorting may serve to equip EVs with distinct biological functions tailored to the hypoxic tumor microenvironment, thus impacting processes like intercellular communication, migration, and invasion. Particularly in our result, approximately 85% of the N-glycans are overlapping in both the cells and EVs, while 27 unique N-glycan patterns were found only in the EVs, predominantly complex fucosylated and sialylated N-glycans. Specifically, N-glycan H7N6F1S2, absent in the MDA-MB-231 cells but present in the EVs, was linked to high diagnostic accuracy in breast cancer [
20]. This highlights the potential of unique EV N-glycan profiles as biomarkers for cancer invasion and metastasis [
18].
Notably, the relative abundance of mono-antennary N-glycans, specifically H4N3F1S2, H3N3F1S0, and H7N4F3S2, was significantly elevated in the hypoxic EVs, suggesting dysregulation or incomplete maturation of N-glycan biosynthesis associated with altered cell cycle dynamics, which reflects the heightened cellular proliferation in hypoxia [
21]. Moreover, complex N-glycans with varied branching and fucose residues indicate intricate N-glycan synthesis pathways influenced by hypoxic stress. Specific N-glycan structures, like H6N4F1S1 and H5N4F1S1, showed significant alterations in the hypoxic EVs, suggesting their involvement in enhancing breast cancer invasiveness and potential as therapeutic targets [
20,
22]. Upregulated N-glycans like H8N4F1S0 and H8N6F1S2 in the hypoxic EVs suggest that these structures support cell–cell communication and promote survival under hypoxic conditions, which is potentially enhanced during cancer progression and metastasis [
11]. Specifically, the significant expression of these N-glycans in both chemoresistant and chemosensitive cell lines under hypoxia underscores their potential involvement in cancer cells’ adaptive responses to hypoxic stress. Targeting the glycosylation pathways may disrupt the adaptive mechanisms of cancer cells in hypoxic tumor microenvironments [
1]. Our correlation analysis between the differentially expressed N-glycans and dysregulated N-glycosylation-related proteins in the cells and EVs showed that the dysregulated N-glycans in the hypoxia-induced EVs display positive and negative correlations with the abundance of specific cellular proteins involved in N-glycosylation synthesis and glycolysis. For instance, the quantitative reduction in sialylation observed in the hypoxia-induced cells and EVs can be attributed to the downregulation of NANS expression, an enzyme crucial for N-acetylmannosamine-6-phosphate to N-acetylmannosamine-9-phosphate, a pivotal step in the biosynthesis of endogenous sialic acid [
23]. Notably, DPM3 emerges as a central component of the dolichol phosphate–mannose (DPM) complex, which is essential for synthesizing DPM, a key precursor for the biosynthesis of high-mannose N-glycans within the endoplasmic reticulum [
24]. Importantly, our findings indicate a strong correlation between DPM3 and the high-mannose N-glycan Man9, thereby emphasizing its crucial role in the glycosylation pathway. Although our analysis implicates correlations between the N-glycan profile and the abundance of proteins involved in the asparagine N-linked glycosylation and glycolysis/gluconeogenesis pathways, the molecular mechanisms must be investigated further.
4. Materials and Methods
4.1. Cell Lines and Cell Culture
The MDA-MB-231 breast cancer cell line was purchased from ATCC (Cell Lines Service, Eppelheim, Germany). MDA-MB-231 cells were cultured in DMEM supplemented with 10% FCS and 2 mM L-glutamine (Merck, Darmstadt, Germany). Cells were maintained at 37 °C in a humidified atmosphere with 5% CO2. Hypoxia treatment was performed at 37 °C in a water-saturated atmosphere at 5% CO2 in the incubator Heracell 15 (Thermo Fisher Scientific, Waltham, MA, USA) at 1% O2. The oxygen partial pressure was adjusted by N2.
4.2. EV Isolation
EV donor cells (MDA-MB-231) were slowly adapted to serum-free DMEM. Culture supernatants (100 mL) from cells cultured for 48 h under normoxia or hypoxia were collected, and protease inhibitors were added. The supernatants were centrifuged for 10 min at 500×
g, followed by another 10 min at 5500×
g to remove cells and cell debris. To isolate EVs, the supernatant was passed through a 0.22 µm and 0.1 µm sodium acetate filter (Cytiva, Freiburg, Germany). EVs were isolated by ultracentrifugation (70 min at 100,000×
g, 4 °C) in an Optima L-100 XP Ultracentrifuge (Beckman Coulter, Brea, CA, USA) using an SW40Ti rotor following the procedure outlined by Santra et al. [
25].
4.3. Nanoparticle Tracking Analysis (NTA)
Cell-culture-derived EVs were diluted with PBS 1:500 or 1:1000. Five videos of 30 s each were recorded (settings: camera level = 16, screen gain = 2) using the NanoSight microscope (LM14, Amesbury, UK). The NanoSight NTA 3.0 software’s processing function was used to analyze the recordings (settings: detection threshold = 6, screen gain = 2). PBS after 0.1 µm sodium acetate filter was also analyzed as background control and showed almost no signal.
4.4. Western Blot Analysis and Antibodies
MDA-MB-231 cells and their derived EVs were lysed at 4 °C using ice-cold sodium deoxycholate (SDC, (Merck, Darmstadt, Germany)) lysis buffer containing 100 mM triethylammonium bicarbonate and 1% (w/v) SDC. Protein extracts were analyzed via SDS-PAGE and Western blotting, with protein concentrations determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Dreieich, Germany). Protein lysates were mixed with NuPAGE LDS Sample Buffer (Thermo Fisher Scientific, Dreieich, Germany), heated for 10 min at 99 °C, and equal amount of protein was resolved in a NuPage 12% Bis Tris gel (Thermo Fisher Scientific, Dreieich, Germany). Following electrophoretic separation, proteins were transferred onto nitrocellulose membranes (Thermo Fisher Scientific, Dreieich, Germany) via wet blotting and blocked for 45 min with RotiBlock (Roth Chemie, Karlsruhe, Germany) in Tris-buffered saline with 0.1% Tween 20 (Merck, Darmstadt, Germany) containing 5% skimmed dry milk (Merck, Darmstadt, Germany) under gentle agitation at room temperature. The membranes were then incubated overnight at 4 °C with primary antibodies (diluted 1:1000), including anti-HIF1α (610958, BD Biosciences, Franklin Lakes, NJ, USA), anti-CD81 (56039T, Cell Signaling Technology, Danvers, MA, USA), anti-flotillin-1 (610820, BD Biosciences, Franklin Lakes, NJ, USA), anti-Lamin C (AB108595, Abcam, Cambridge, UK), and β-actin (4967S, Cell Signaling Technology, Danvers, MA, USA). After washing, the membranes were incubated with HRP-conjugated anti-mouse IgG (1:10,000) (LI-COR Biosciences, Lincoln, NE, USA) or anti-rabbit secondary antibody (1:3000) (LI-COR Biosciences, Lincoln, NE, USA). Protein bands were visualized using SuperSignal West Femto substrate (Thermo Fisher Scientific, Dreieich, Germany) with a ChemiDoc imaging station (BioRad, Feldkirchen, Germany), and densitometric quantification was performed using Quantity One software (Version 4.6.8, BioRad, Feldkirchen, Germany).
4.5. Mass Spectrometry-Based Proteomics and N-Glycomics
4.5.1. N-Glycan Release and Peptide Preparation
Further, 50 µg of protein per cell lysate sample and 40 μg of protein per EV lysate sample were both diluted in SDC lysis buffer (0.1 M TEAB with 1% SDC) to a final volume of 100 µL. Subsequently, the samples were reduced and alkylated with 10 mM dithiothreitol (DTT, Sigma Aldrich, St. Louis, MO, USA) for 30 min at 56 °C and 20 mM iodoacetamide (IAA, Sigma Aldrich) for 30 min at 37 °C in the dark. Concentration and buffer exchange were performed using amicon ultra centrifugal filters filled with 100 mM ammonium bicarbonate. Protein digestion was performed by adding trypsin to each sample at a ratio of 1:100 and incubating at 37 °C for 18 h. N-glycans were cleaved from peptides using thirty units of PNGase F followed by enzymatic processing. The samples were vacuum-dried and then resuspended in 200 μL of 5% (v/v) acetic acid for the purification of N-glycans using the Sep-Pak C18 cartridge. Native N-glycan and peptide samples were dried using a SpeedVac concentrator (Thermo Fisher Scientific, Dreieich, Germany).
4.5.2. N-Glycan Reduction and Solid-Phase Permethylation
N-glycan reduction and permethylation were performed following Guan et al. [
26]. Released N-glycans were treated with a borane–ammonia complex at 60 °C for 1.5 h to remove α and β anomers from the reducing end. The reduced N-glycans were subsequently dried and re-dissolved in methanol to eliminate residual ammonium hydroxide and ammonium carbonate, drying using a SpeedVac concentrator. The permethylation of N-glycans was then performed using an optimized solid-phase method. Briefly, N-glycans were dissolved in DMSO/water solution, combined with methyl iodide, and incubated with NaOH beads at room temperature. The permethylation and any subsequent oxidation reactions were quenched with 5% acetic acid, and permethylated N-glycans were extracted using a chloroform–water separation method. The chloroform phases containing the permethylated N-glycans were vacuum-dried and stored at −40 °C for subsequent analysis or measurement.
4.5.3. Peptide Purification by Single-Pot Solid-Phase-Enhanced Sample Preparation (SP3)
The dried peptides were purified using the SP3 protocol, as described by Hughes, C.S., et al. [
27]. Briefly, the dried peptides were resuspended in 95% ACN, and 5 μL of SP3 beads (magnetic beads) were followed by vortexing and settling on the magnetic rack. The sample was washed twice with 500 μL of ACN, and the peptides were eluted with 50 μL of 2% DMSO/1% FA. The elution was centrifuged at 16,000×
g for 5 min, and the supernatant was moved to a new tube. The samples were dried in a vacuum microcentrifuge concentrator.
4.5.4. Analysis of N-Glycans and Peptides on Nano-LC–ESI-MS/MS
Analysis of reduced permethylated N-glycans and label-free peptides was performed using mass spectrometry under the same conditions as previously described by Godbole S et al. [
28]. Briefly, chromatographic separation of peptides and N-glycans was achieved with a two-buffer system (buffer A: 0.1% FA in H
2O; buffer B: 0.1% FA in ACN) on a nano-UHPLC (Dionex Ultimate 3000 UHPLC system, Thermo Fisher). Attached to the UHPLC was a peptide trap (100 µm × 20 mm, 100 Å pore size, 5 µm particle size, C18, Nano Viper, Thermo Fisher) for online desalting and purification, followed by a 25 cm C18 reversed-phase column (75 µm × 250 mm, 130 Å pore size, 1.7 µm particle size, peptide BEH C18, nanoEase, Waters). Sample measurement was performed in randomized order, and injection of 100 ng HeLa protein digest standard (88329, Thermo Fisher Scientific) was used for quality control before and after measurements.
Peptides were separated using an 80 min method with linearly increasing ACN concentration from 2% to 30% ACN over 60 min. MS/MS measurements were performed on a quadrupole-ion-trap-orbitrap MS (Orbitrap Fusion, Thermo Fisher). Eluting peptides were ionized using a nano-electrospray ionization source (nano-ESI) with a spray voltage of 1800 V and analyzed in data-dependent acquisition (DDA) mode. For each MS1 scan, ions were accumulated for a maximum of 120 milliseconds or until a charge density of 2 × 105 ions (AGC Target) was reached. Fourier-transformation-based mass analysis of the data from the orbitrap mass analyzer was performed covering a mass range of m/z 400–1300 with a resolution of 120,000 at m/z = 200. Peptides with charge states between 2+ and 5+ above an intensity threshold of 1000 were isolated within a m/z 1.6 isolation window in Top Speed mode for 3 s from each precursor scan and fragmented with a normalized collision energy of 30% using higher-energy collisional dissociation (HCD). MS2 scanning was performed using an ion trap mass analyzer at a rapid scan rate, covering a mass range starting at m/z 120 and accumulated for 60 ms or to an AGC target of 1 × 105. Already fragmented peptides were excluded for 30 s.
For permethylated N-glycan analysis, a 115 min gradient was utilized, beginning with 2% solvent B, which increased to 30% over 10 min, followed by an increase to 75% over 70 min, and concluding at 95% by the end of the gradient. N-glycans were analyzed using a quadrupole–orbitrap–ion trap mass spectrometer in DDA mode. For MS1 scanning, an orbitrap mass analyzer with a resolution of 120,000 FWHM was used, with an AGC target of 2.0 × 10
6 and an
m/
z scan range from 400 to 2000. For collision-induced dissociation (CID)-MS/MS, the most intense precursor ions were selected for fragmentation, isolated using a 2 m/z window, with a normalized collision energy of 35%. Fragments were detected at an orbitrap resolution of 17,500 FWHM, with an AGC target of 1.0 × 10
4 and a maximum accumulation time of 20.00 ms. MS data are available via ProteomeXchange [
29] with identifier PXD053696 and GlycoPOST [
30] with identifier GPST000466.
4.6. Processing of Proteome and N-Glycome Raw Data
Proteome raw data were searched with the Sequest algorithm integrated into the Proteome Discoverer (Version 3.0.0.757, Thermo Fisher Scientific) against a reviewed human Swissprot FASTA database, obtained in April 2021, containing 20,365 entries. Subsequent data processing and proteome analysis were conducted using Perseus (version 2.0.11.0).
Raw N-glycan data visualization was carried out with Xcalibur Qual Browser, and molecular masses were extracted utilizing MaxQuant. Potential monosaccharide compositions of N-glycan precursors were determined through an in-house Python script [
26]. The identification of N-glycan compositions and structures was based on MS/MS results, with manual assignments correlating observed peaks in MS/MS spectra to potential N-glycan fragments listed in GlycoWorkbench 2.1. N-glycan compositions were annotated with specific symbols representing monosaccharide components (H (Hexose)-N (HexNAc + red-HexNAc)-F (Fucose)-S (Neu5Ac)). The determined N-glycan structures adhered to the Symbol Nomenclature for Glycans guidelines and were visually represented using GlycoWorkbench [
31,
32]. Quantitative analysis of N-glycans was facilitated using Skyline software 21.1.0.278, with peak area data subsequently exported for further analysis in Excel.
4.7. Statistical Analysis and Visualization
Relative protein abundances were log2-transformed and normalized to the median protein abundance in each sample using Perseus to correct for variations in sample-to-sample injected amounts. Protein-enriched pathways were assessed by the Genomes (KEGG) database and Reactome Knowledgebase [
33] and the clusterProfiler R package 4.12.6 [
34], with significant pathways identified at a
p-value < 0.05. Identified EV proteins were compared with data from the ExoCarta database, and proteomic profiles of EVs were analyzed using Metascape for pathway enrichment [
35]. GSEA was employed to identify and enrich protein sets related to “Asparagine N-linked glycosylation” and “Glycolysis/Gluconeogenesis” from cell proteome profiles. The functional interaction among significant expressed Asparagine N-linked glycosylation proteins was derived from the STRING database [
36]. Additionally, MetScape was employed for the integration and visualization of dysregulated metabolites alongside dysregulated proteins in the glycolysis/gluconeogenesis pathways to map interactions and biological pathways impacted in hypoxia-induced cells. Relative quantification of individual N-glycan species was performed by calculating their percentages relative to the total N-glycans detected, with traits such as truncation, high-mannose, hybrid-type, complex-type, number of antennas, bisecting GlcNAc, sialylation, fucosylation, galactosylation, poly-LacNAc motifs, Lewis A/X, and sLewis A/X analyzed [
37]. Statistical assessments of significant changes in glycosylation features were performed using one-way ANOVA, and dysregulations in proteins and N-glycans were identified using the Student’s
t-test with the Benjamini–Hochberg FDR adjustment method (
p-value < 0.05; fold change > 1.5;
q-value < 0.05). The nine-quadrant map analysis and Circos plot were both utilized to identify significant correlations between differentially expressed N-glycans in hypoxic vs. normoxic cells and hypoxic vs. normoxic EVs, and dysregulated N-glycosylation-related proteins using Pearson’s correlation coefficients. PLS-DA was employed to differentiate sample cohorts based on proteins and N-glycans. Data visualization, including volcano plots, heatmaps, Venn diagrams, and scatter plots, was performed using R 4.2.2 and GraphPad Prism software (Version 7.0), with results considered statistically significant at
p-value < 0.05.