Global extracellular vesicle proteomic signature defines U87-MG glioma cell hypoxic status with potential implications for non-invasive diagnostics
Glioblastoma multiforme (GBM) is the most common and lethal of primary malignant brain tumors. Hypoxia constitutes a major determining factor for the poor prognosis of high-grade glioma patients, and is known to contribute to the development of treatment resistance. Therefore, new strategies to comprehensively profile and monitor the hypoxic status of gliomas are of high clinical relevance. Here, we have explored how the proteome of secreted extracellular vesicles (EVs) at the global level may reflect hypoxic glioma cells.
We have employed shotgun proteomics and label free quantification to profile EVs isolated from human high-grade glioma U87-MG cells cultured at normoxia or hypoxia. Parallel reaction monitoring was used to quantify the identified, hypoxia-associated EV proteins. To determine the potential biological significance of hypoxia-associated proteins, the cumulative Z score of identified EV proteins was compared with GBM subtypes from HGCC and TCGA databases.
In total, 2928 proteins were identified in EVs, out of which 1654 proteins overlapped with the ExoCarta EV-specific database. We found 1034 proteins in EVs that were unique to the hypoxic status of U87-MG cells. We subsequently identified an EV protein signature, “HYPSIGNATURE”, encompassing nine proteins that strongly represented the hypoxic situation and exhibited close proximity to the mesenchymal GBM subtype.
We propose, for the first time, an EV protein signature that could comprehensively reflect the hypoxic status of high-grade glioma cells. The presented data provide proof-of-concept for targeted proteomic profiling of glioma derived EVs, which should motivate future studies exploring its utility in non-invasive diagnosis and monitoring of brain tumor patients.
KeywordsGlioblastoma Hypoxia Extracellular vesicles Mass spectrometry Label free quantification
Glioblastoma multiforme (GBM) is the most common and malignant type of primary brain tumor in adults with a median survival of approximately 15 months [1, 2, 3]. GBM is identified from less malignant, low grade gliomas, by extensive regions of hypoxia  that directly correlate with the aggressive behaviour . Hypoxia results from the high proliferative and metabolic activity of malignant cells  and is associated with pseudopalisading necrosis as well as vascular hyperproliferation . Tumor hypoxia modulates stromal cell interactions in the microenvironment that further support the survival and dissemination of malignant cells [4, 8, 9, 10, 11]. Numerous studies have previously shown that tumor progression is driven by hypoxic signaling , and the expression of hypoxia-related markers correlate with poor patient outcome in several tumor types, including GBM . However, the development of strategies for non-invasive monitoring of brain tumor hypoxic signalling remains a challenge of high clinical relevance, especially with regard to the relative inaccessibility and spatiotemporal heterogeneity of GBM tumors.
Extracellular vesicles (EVs) are excessively secreted by tumor cells into the circulation, and are emerging as a promising candidate for liquid biopsy-based approaches in cancer [14, 15, 16]. Exosomes and microvesicles are lipid-bilayer EVs  that have come to be recognized in intercellular communication, promoting the development and progression of various disease conditions . Numerous studies have shown that exosome-like EVs may mediate hypoxia-dependent intercellular signaling in GBM . Moreover, pilot studies based on an antibody array targeted at angiogenesis-related proteins, suggested that the EV proteome may reflect the tumor oxygenation status in GBM . To further develop EV-based strategies for non-invasive tumor diagnosis and monitoring of hypoxia, it is essential to comprehensively identify proteins that are efficiently sorted to EVs and that reflect the hypoxic status of the cell or tissue of origin.
In this study, we employed label free quantification (nontargeted method) and parallel reaction monitoring (targeted method) to globally characterize the proteome of EVs derived from U87-MG high-grade glioma cells with the aim to understand how EV profiling can be exploited to noninvasively define the hypoxic status of glioma tumors.
Global proteome identification in EVs derived from high-grade glioma cells
We then employed shotgun proteomics by data-dependent acquisition to comprehensively determine the proteome of EVNORM and EVHYP derived from U87-MG cells. We identified a total of 2089 EVHYP and 2035 EVNORM proteins (Fig. 1f; Supplementary Tables 1, 2). There were 1034 protein groups unique to EVHYP (Fig. 1f; Supplementary Table 3) and 1055 protein groups common to both EVNORM and EVHYP (Fig. 1f; Supplementary Table 4). We next created a multiconsensus list combining EVNORM and EVHYP protein identities (Supplementary Table 5) and then compared the multiconsensus protein group to the ExoCarta EV public database . The multiconsensus EV identities (2928 proteins) showed extensive overlapping of 1654 common identities with the ExoCarta database and also identified 1274 unique identities (Fig. 1g), which support the sensitivity of detection of the EV proteome with the current approach.
Processing of the EV proteome by label free quantification (LFQ)
Hsignificant proteins were found to be distributed mostly in nucleic acid binding, hydrolase, enzyme modulators, and cytoskeletal protein subclasses, as determined by Gene Ontology system of classification using PANTHER version 14.0  (Fig. 2b). Then we analysed the differences in functional classification pertaining to biological processes, molecular functions, and cellular localization of the Hsignificant (orange bars) and Hnonsignificant (blue bars) proteins (Fig. 2). A substantially higher number of Hsignificant proteins were localized in organelles (GO:0,043,226) and macromolecular complexes (GO:0,032,991) compared to Hnonsignificant proteins (Fig. 2c). More Hsignificant proteins were associated with cellular (GO:0,009,987), metabolic (GO:0,008,152), and cellular component biogenesis processes (GO:0,071,840) and catalytic activity (GO:0,003,824) compared to Hnonsignificant proteins (Fig. 2d, e), consistent with characteristics of the hypoxic tumor state [30, 31].
Validation of Hsignificant profile by parallel reaction monitoring (PRM)
The peak normalized areas (PAN) of individual peptides of all proteins analyzed were extracted from the Skyline, and the average of replicate PAN values of each individual peptide of all proteins in EVNORM or EVHYP samples were calculated. In all cases, the selected peptides of 135 candidate proteins had quantifiable distribution of area under curve for the identified peptide transitions (Supplementary Table 8). On analysing the fold change, we found 17 proteins significantly differentially expressed in EVHYP as compared to EVNORM (Fig. 3b; Supplementary Table 8). We further applied peptide significance and normalized peak area restrictions on the hypoxia response of the Hsignificant EV proteins (N = 17) and filtered it down to a signature of 9 proteins that included Insulin-like Growth Factor-Binding Protein 3 (IGFBP3), Tissue Factor (F3), Carbonic Anhydrase 9 (CA9), Solute Carrier Family 2 Facilitated Glucose Transporter Member 1 (SLC2A1), Nucleolin (NCL), Osteopontin (SPP1), Monocarboxylate Transporter 1 (SLC16A1), Membrane-Associated Progesterone Receptor Component 1 (PGRMC1), and Annexin A5 (ANXA5) (Fig. 3c). These proteins defined a profile of unique proteins (N = 9) efficiently sorted from donor cells to EVs and enriched at hypoxic conditions, hereafter referred to as “HYPSIGNATURE” (the PAN of the replicates of the different peptides is given in Supplementary Fig. 1).
HYPsignature can identify GBM mesenchymal subtype
In this study, we used an optimized combination of nontargeted and targeted quantitative proteomics to comprehensively profile hypoxia-regulated proteins associated with high-grade glioma cell derived EVs. We have identified a protein signature, “HYPSIGNATURE”, in EVs secreted by U87-MG cells that is associated with the HIF hypoxic signaling response and exhibited close proximity to the mesenchymal GBM subtype. Importantly, out of the nine proteins encompassing the HYPSIGNATURE, seven proteins are known as plasma membrane integrated proteins with an extracellular domain available for specific recognition by antibodies and other targeting agents. Together, our findings thus propose that the hypoxic status of GBM tumors can be defined by the EV HYPSIGNATURE, which may be utilized not only to noninvasively immunephenotype glioma tumors but also as potential therapeutic targets.
The utility of EVs across diverse cellular functions, including recent investigations that support the application of EVs as non-invasive biomarker tools [14, 16, 37, 38], strongly motivates improved efforts to comprehensively profile the proteome of EVs derived from cells grown at disease mimicking conditions. Using discovery proteomics, a previous study  identified a total of 844 proteins in EVs isolated from GBM cells. In comparison, we identified approximately 3000 proteins in EVs, out of which 1034 proteins were unique to hypoxic EVs. Importantly, the major aim of the present study was to specifically identify an EV signature that mimics the hypoxic situation, i.e. a pathognomonic feature of GBM tumors associated with disease aggressiveness and treatment resistance. Although the studies are limited to one glioma cell-line, it may be argued that the obtained results have general relevance given the substantial overlap between EV protein identities found here and the ExoCarta EV proteome database. Moreover, the hypoxic response is a universal phenomenon of high-grade gliomas as well as other highly malignant tumors. Clearly, future studies will have to further assess the generalizability of the present data, including validation in primary GBM cell models as well as in vivo.
LFQ has now become a widely accepted analytical approach for comparison of the relative abundance of proteins across multiple samples [40, 41, 42]. The possibility to analyse untreated proteins or peptides in a large number of samples makes LFQ a preferred protocol over other relative quantification approaches. However, previous studies have shown that sample preparation for the LFQ approach is highly susceptible to variability . Therefore, to reduce this variability, we used 9 replicates of normoxia and 12 replicates of hypoxia samples for LFQ. In addition, the conforming pattern of differential levels of most proteins analyzed in LFQ (Supplementary Table 6) and PRM (Supplementary Table 8), suggest a high degree of sample preparation consistency. In support of EV proteomics data, immunoblotting showed an enrichment of top candidates of the HYPSIGNATURE, and gene array analysis showed increased expression of IGFBP3, F3, CA9, SLC2A1 and PGRMC1 mRNA in hypoxic as compared with normoxic U87-MG cells. We were unable to detect other candidate proteins (NCL, SLC16A1, SPP1, ANXA5) in EVs by immunoblotting analysis, either from normoxia or hypoxia, and did not detect a hypoxic enrichment of these proteins in U87-MG cells. A potential explanation to the discrepancy between an induction of these proteins in EVs collected over a cumulative time period of 48 h of hypoxia, and cells analyzed at a fixed time-point, is the well-known temporal dynamics of the hypoxic response.
Several previous studies have associated tumor cell expression of HYPSIGNATURE proteins with increased GBM aggressiveness. For example, F3 expression was demonstrated to be hypoxia-dependent in highly aggressive P7 GBM cells, leading to increased F3 activity , and F3-positive EVs were shown to induce angiogenesis . Hypoxia also induced increased SLC16A1 plasma membrane expression in glioma cells, both in in vitro and in vivo models . Additionally, SLC16A1 plasma membrane expression was associated with HIF-1α and CA9 positivity in hypoxic regions. Further, SLC16A1 was found to be upregulated in GBM as compared with normal tissues . NCL was also found to be overexpressed in patient-derived GBM tumors and cells as compared with normal brain . ANXA5 has been found to promote invasion and chemoresistance to the alkylating drug temozolomide in GBM cells . Since hypoxic cells and components in the hypoxic niche have been increasingly implicated in resistance to temozolomide , it is conceivable that ANXA5 is associated with the hypoxic component of drug resistance. SPP1 was shown to be induced by hypoxia both in vitro and in vivo  and is predominantly observed in the microvasculature of GBM . Several studies have implicated SPP1 with crucial roles in invasion  and malignant gliomas . In several glioma cell models, CA9 strongly co-localized with HIF-1α, indicating its induction in hypoxic regions of this tumor type. Clinically, CA9 is minimally expressed in normal brain tissue, whereas its high expression in brain tumors strongly correlated with the level of malignancy . SLC2A1 is another well-established hypoxia-induced protein that has been associated with hypoxic regions of GBM . These studies support a functional role of HYPSIGNATURE protein expression in tumor cells, and future studies that define the tumor promoting role of these proteins when associated with EVs, especially in the context of e.g. pH regulation (CA9), metabolite transport (SLC2A1, SLC16A1), and coagulation activation (F3), will be of high interest.
To conclude, our data strongly support that a specific subset of mostly membrane intercalated EV proteins could define the hypoxic status of high-grade glioma cells. The proteins identified as part of the HYPSIGNATURE warrant further clinical examination using a targeted approach to validate their capacity to differentiate the highly heterogeneous nature of high-grade glioma tumors from e.g. low grade gliomas and other brain lesions that are challenging to define by imaging alone. This proof-of-principle study to noninvasively define the glioma hypoxic status utilizing advanced proteomics is a significant step in this direction.
Materials and methods
U87-MG cells were newly purchased from ATCC. Cells were routinely cultured in DMEM medium, supplemented with 10% foetal bovine serum (FBS), 2 mM L-glutamine, 100 U/mL penicillin and 100 μg/mL streptomycin (growth medium). All cells were grown in humidified 5% CO2 incubator at 37 °C. For hypoxia experiments, cells were incubated in humidified Sci-tive NN Hypoxia workstation (Ruskinn Technology) set at 5% CO2, 1% O2, and 37 °C.
Normoxic or hypoxic EVs were isolated in parallel from U87-MG cells at a particular passage by standard procedures, using differential ultracentrifugation . Routinely cultured U87-MG cells at sub-confluency were grown in DMEM supplemented with 1% BSA at normoxic or hypoxic conditions for 48 h. Conditioned media were collected after 48 h and centrifuged at 300×g twice to eliminate cell debris. Supernatant fractions were then centrifuged at 100,000×g for 2 h to pellet EVs, followed by washing twice with PBS at 100,000×g for 2 h. EVs were then resuspended in 6 M Urea for downstream proteomics experiments.
U87-MG cells or EV protein lysate were mixed with NuPAGE 4 × LDS Sample Buffer (Life Technologies) and heated for 10 min at 80 °C. Equal amount of proteins was resolved in a NuPage 4–12% Bis Tris gel (Life Technologies) at non-reducing or reducing conditions and then transferred onto a polyvinylidene fluoride (PVDF) membrane (Immobilon-FL), followed by blocking in TBS containing 0.05% Tween 20, 5% nonfat dry milk or 3% BSA for 1 h at RT. To probe for CD9, CD63, CD81, Flotillin-1, IGFBP3, and CA9, the membrane was incubated with the following antibodies in TBST containing 5% nonfat dry milk overnight at 4 °C: anti-CD9 (1:2000; ab92726, Abcam), anti-CD63 (1:100; ab8219, Abcam), anti-CD81 (1:1000; ab109201, Abcam), anti-flotillin-1 (1 µg/mL; ab41927, Abcam), Rabbit anti-IGFBP3 (1:80; PAAJ1, GroPep), M75 anti-CA9 (1:300; M75, Bioscience Slovakia), Mouse anti-NCL (1:1000, ab13541, Abcam), Rabbit anti-SLC16A1 (1:1000, ab179832, Abcam), Mouse anti-SPP1 (1:500, ab166709, Abcam), and Rabbit anti-ANXA5 (1:500, ab14196, Abcam). After washing, the membrane was incubated with HRP-conjugated anti-mouse IgG (1:10,000) (A9044, Sigma-Aldrich) or anti-rabbit secondary antibody (1:3000) (7074, Cell Signaling Technology). Protein bands were visualized by enhanced chemiluminescence western blotting substrate (Pierce).
Nanoparticle Tracking Analysis, Transmission Electron Microscopy, Trypsin digestion and peptide preparation, Discovery LC–MS/MS, label free quantification, and quantitative LC-PRM-MS/MS were performed as described in Supplementary Materials and Methods.
The Gene Ontology functional classification of Hsignificant proteins was performed using PANTHER (https://www.pantherdb.org/). Enriched pathways of EVHYP signature proteins were determined using ConsensusPathDB-human interaction database (https://cpdb.molgen.mpg.de/). Wilcoxon test was employed for pathway enrichment analysis with a P value cut-off of 0.01.
Gene expression data on different GBM subtypes were obtained from The Cancer Genome Atlas (TCGA) via the GlioVis portal (https://gliovis.bioinfo.cnio.es/), as well as from the Human Glioma Cell Cultures (HGCC) database (https://www.hgcc.se/).
For HGCC data analysis, the gene expression Z score for each HYPSIGNATURE candidate in subtypes (Classical, Mesenchymal, Proneural, or Neural) was directly extracted from the HGCC database. Cumulative Z score was generated as described for TCGA dataset.
Data are expressed as mean ± STDEV. Statistical analyses were done using unpaired Student t test. All values with P < 0.05 were considered to be statistically significant.
Open access funding provided by Lund University. We thank Maria Johansson and Melinda Rezeli for their advice with parallel reaction monitoring protocol optimization. This study was funded by grants (to MB) from the Swedish Cancer Fund; the Swedish Research Council; the Swedish Childhood Cancer Foundation; the Gunnar Nilsson Cancer Foundation; the Mrs Berta Kamprad Foundations; the Skåne University Hospital donation funds; the Governmental funding of clinical research within the national health services, ALF; and a donation by Viveca Jeppsson. Strategic infrastructure support from ThermoFisher Scientific (to GMV) is also acknowledged.
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflicts of interest.
- 3.Arvold ND, Reardon DA (2014) Treatment options and outcomes for glioblastoma in the elderly patient. Clin Interv Aging 9:357–367Google Scholar
- 14.Chandran VI, Welinder C, Mansson AS, Offer S, Freyhult E, Pernemalm M, Lund SM, Pedersen S, Lehtio J, Marko-Varga G, Johansson MC, Englund EM, Sundgren PC, Belting M (2019) Ultrasensitive immunoprofiling of plasma extracellular vesicles identifies syndecan-1 as a potential tool for minimally invasive diagnosis of glioma. Clin Cancer Res. https://doi.org/10.1158/1078-0432.CCR-18-2946.Google Scholar
- 19.Bang-Rudenstam A, Cerezo-Magana M, and Belting M (2019) Pro-metastatic functions of lipoproteins and extracellular vesicles in the acidic tumor microenvironment. Cancer Metastasis RevGoogle Scholar
- 20.Svensson KJ, Kucharzewska P, Christianson HC, Skold S, Lofstedt T, Johansson MC, Morgelin M, Bengzon J, Ruf W, Belting M (2011) Hypoxia triggers a proangiogenic pathway involving cancer cell microvesicles and PAR-2-mediated heparin-binding EGF signaling in endothelial cells. Proc Natl Acad Sci USA 108:13147–13152CrossRefGoogle Scholar
- 21.Drucker KL, Kitange GJ, Kollmeyer TM, Law ME, Passe S, Rynearson AL, Blair H, Soderberg CL, Morlan BW, Ballman KV, Giannini C, Jenkins RB (2009) Characterization and gene expression profiling in glioma cell lines with deletion of chromosome 19 before and after microcell-mediated restoration of normal human chromosome 19. Genes Chromosomes Cancer 48:854–864CrossRefGoogle Scholar
- 22.Jung T-Y, Choi Y-D, Kim Y-H, Lee J-J, Kim H-S, Kim J-S, Kim S-K, Jung S, Cho D (2013) Immunological characterization of glioblastoma cells for immunotherapy. Anticancer Res 33:2525–2533Google Scholar
- 30.DeBerardinis RJ, Chandel NS (2016) Fundamentals of cancer metabolism Science advances 2:e1600200–e1600200Google Scholar
- 34.Phillips HS, Kharbanda S, Chen R, Forrest WF, Soriano RH, Wu TD, Misra A, Nigro JM, Colman H, Soroceanu L, Williams PM, Modrusan Z, Feuerstein BG, Aldape K (2006) Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell 9:157–173CrossRefGoogle Scholar
- 35.Verhaak RG, Hoadley KA, Purdom E, Wang V, Qi Y, Wilkerson MD, Miller CR, Ding L, Golub T, Mesirov JP, Alexe G, Lawrence M, O'Kelly M, Tamayo P, Weir BA, Gabriel S, Winckler W, Gupta S, Jakkula L, Feiler HS, Hodgson JG, James CD, Sarkaria JN, Brennan C, Kahn A, Spellman PT, Wilson RK, Speed TP, Gray JW, Meyerson M, Getz G, Perou CM, Hayes DN (2010) Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 17:98–110CrossRefGoogle Scholar
- 37.Yang KS, Im H, Hong S, Pergolini I, Del Castillo AF, Wang R, Clardy S, Huang CH, Pille C, Ferrone S, Yang R, Castro CM, Lee H, Del Castillo CF, Weissleder R (2017) Multiparametric plasma EV profiling facilitates diagnosis of pancreatic malignancy. Sci Transl Med 9(391):eaal3226CrossRefGoogle Scholar
- 38.Figueroa JM, Skog J, Akers J, Li H, Komotar R, Jensen R, Ringel F, Yang I, Kalkanis S, Thompson R, LoGuidice L, Berghoff E, Parsa A, Liau L, Curry W, Cahill D, Bettegowda C, Lang FF, Chiocca EA, Henson J, Kim R, Breakefield X, Chen C, Messer K, Hochberg F, Carter BS (2017) Detection of wild-type EGFR amplification and EGFRvIII mutation in CSF-derived extracellular vesicles of glioblastoma patients. Neuro Oncol 19:1494–1502CrossRefGoogle Scholar
- 45.Miranda-Goncalves V, Granja S, Martinho O, Honavar M, Pojo M, Costa BM, Pires MM, Pinheiro C, Cordeiro M, Bebiano G, Costa P, Reis RM, Baltazar F (2016) Hypoxia-mediated upregulation of MCT1 expression supports the glycolytic phenotype of glioblastomas. Oncotarget 7:46335–46353CrossRefGoogle Scholar
- 47.Balça-Silva J, do Carmo A, Tão H, Rebelo O, Barbosa M, Moura-Neto V, Sarmento-Ribeiro AB, Lopes MC, Moreira JN (2018) Nucleolin is expressed in patient-derived samples and glioblastoma cells, enabling improved intracellular drug delivery and cytotoxicity. Exp Cell Res 370:68–77CrossRefGoogle Scholar
- 55.Joseph JV, Conroy S, Pavlov K, Sontakke P, Tomar T, Eggens-Meijer E, Balasubramaniyan V, Wagemakers M, den Dunnen WFA, Kruyt FAE (2015) Hypoxia enhances migration and invasion in glioblastoma by promoting a mesenchymal shift mediated by the HIF1α–ZEB1 axis. Cancer Lett 359:107–116CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.