Extracellular Vesicles from Neurosurgical Aspirates Identifies Chaperonin Containing TCP1 Subunit 6A as a Potential Glioblastoma Biomarker with Prognostic Significance

Glioblastoma, WHO‐grade IV glioma, carries a dismal prognosis owing to its infiltrative growth and limited treatment options. Glioblastoma‐derived extracellular vesicles (EVs; 30–1000 nm membranous particles) influence the microenvironment to mediate tumor aggressiveness and carry oncogenic cargo across the blood–brain barrier into the circulation. As such, EVs are biomarker reservoirs with enormous potential for assessing glioblastoma tumors in situ. Neurosurgical aspirates are rich sources of EVs, isolated directly from glioma microenvironments. EV proteomes enriched from glioblastoma (n = 15) and glioma grade II–III (n = 7) aspirates are compared and 298 differentially‐abundant proteins (p‐value < 0.00496) are identified using quantitative LC–MS/MS. Along with previously reported glioblastoma‐associated biomarkers, levels of all eight subunits of the key molecular chaperone, T‐complex protein 1 Ring complex (TRiC), are higher in glioblastoma‐EVs, including CCT2, CCT3, CCT5, CCT6A, CCT7, and TCP1 (p < 0.00496). Analogous increases in TRiC transcript levels and DNA copy numbers are detected in silico; CCT6A has the greatest induction of expression and amplification in glioblastoma and shows a negative association with survival (p = 0.006). CCT6A is co‐localized with EGFR at 7p11.2, with a strong tendency for co‐amplification (p < 0.001). Immunohistochemistry corroborates the CCT6A proteomics measurements and indicated a potential link between EGFR and CCT6A tissue expression. Putative EV‐biomarkers described here should be further assessed in peripheral blood.


Introduction
Gliomas are primary tumors of the central nervous system (CNS). In 2016, the World Health Organization (WHO) updated their classification to group all diffusely infiltrating tumors into a single category. This grouping includes tumors of distinct phenotypic lineages, such as astrocytomas and oligodendrogliomas, on the basis of shared prognostic markers and clinical management. [1] Within this scheme, grade IV glioblastoma (GBM) is the most common, and also the most lethal type. By comparison, lower grade (II-III) gliomas affect younger patients, are less aggressive and are typically characterized by isocitrate dehydrogenase (IDH) mutations, which confer a more favorable prognosis. [2] GBMs are almost universally fatal within a few years, owing to highly infiltrative growth patterns, intratumoral heterogeneity, high recurrence rates, and limited treatment options. This is compounded by a lack of effective clinical monitoring. Their intracranial location makes them inaccessible to routine biopsy, [3] and they rarely metastasize beyond the CNS, despite often disrupting the blood-brain barrier (BBB). [4] Reliable radiographic monitoring is also confounded by pseudo-progression and pseudo-response, which give false impressions of tumor progression and response to therapy, respectively. [5] As such, there is a real need for readily accessible and clinically informative biomarkers that can indicate disease burden and inform treatment decisions. Liquid biopsies, which involve the sampling of tumor-derived molecules from biofluids, i.e., blood, urine, cerebrospinal fluid, and saliva, may help to obviate this, allowing more confident initiation of treatment and delivering improved patient outcomes.
Extracellular vesicles (EVs) are membrane bound nanoparticles secreted by all cells, ranging from 30-1000 nm in diameter. [6,7] The current nomenclature uses EVs as a broad term that encompasses two main classes of 30-1000 nm secreted vesicles: larger 'microvesicles' that form by budding off the outer cell membrane and smaller 'exosomes' that are secreted through endosomal compartments. [8] Their cargo consists of a variety of proteins, nucleic acids, and other biomolecules that are reflective of the identity and molecular state of the originating cell. [9] Tumor cells release EVs in greater quantities than normal cells, [10] thereby providing access to tumor-derived material that can be detected above lower levels of potential 'noise' from nonrelevant tissues. Importantly, EVs can cross the BBB, [11] making them excellent candidates for monitoring brain tumors in situ. [12] EVs also play integral roles in cancer biology [13] ; glioma-derived EVs are key mediators of cell proliferation and angiogenesis. [12] Tumors can respond to environmental stimuli, such as hypoxia, via intercellular EV-signaling, affecting neighboring tumor cells as well as influencing healthy tissue and hijacking immune cells to support glioma expansion and invasion. [14] The feasibility of EVs as disease biomarkers is of increasing interest [15] and this is especially the case following the detection of GBM-derived RNA protected within the double phospholipid membranes of EVs in the circulation. [12,16] In the context of disease monitoring and management, EVs have been shown to be capable of indicating treatment relevant changes in tumor behavior, such as the development of resistance to chemotherapy [17] as well as radiotherapy, [18] the latter of which was shown to induce tumor cell migration in an EV-mediated manner. [19] Moreover, EV liquid biopsies raise the possibility of developing a way to classify and grade glioma tumors pre-operatively. EVs captured from the cerebrospinal fluid of glioma patients were shown to contain mutant IDH1 transcripts, [20] a genetic alteration that is central to WHO 2016 glioma classification. [21] While gliomas can be graded via imaging, the process is imperfect [22] and the potential for proceeding with unnecessary therapy or failing to administer therapy exists.
To date, the vast majority of profiling studies have focused on EVs isolated from blood. [23] Although the ability to detect appropriate biomarkers in biofluids is the sine qua non of a liquid biopsy, blood-derived EVs pose two key obstacles for biomarker discovery. First, EVs in the circulation are highly heterogeneous, released by all body organs with significant proportions being platelet-and endothelial-derived. [6,24] Second, while tumor cells do release more EVs than normal cells, [10] tumor-derived EVs would still be a relatively minor population in the blood [25] ; high enough for targeted detection, but potentially inadequate for biomarker discovery. [26] This necessitates enrichment steps, [27]

Significance Statement
Glioblastoma is a lethal primary brain tumor with extremely limited treatment; after initial surgery and combined radio-and chemotherapy, tumors inevitably recur, at which point they are rapidly fatal. Of paramount importance in efforts to develop and personalize new treatments for glioblastoma is the means to noninvasively monitor tumor molecules in situ. As such there is a critical need for biomarkers that can measure disease burden and treatment response in glioblastoma patients in a safe, accurate, and timely manner, preferably before changes become clinically apparent. Panels of well-defined biomarkers would allow regular monitoring of glioblastoma tumors that are otherwise difficult to biopsy. Neurosurgical aspirates, rich in tumor tissue and viable cells, were used for harvesting clinical EVs analyzed here. These specimens are collected during tumor resection, which is a critical time point for biomarker discovery as it is the only time at which the tumor microenvironment is directly accessible. If developed and implemented alongside new treatments, such biomarkers would be useful surrogate endpoints and allow clinical trial protocols to be more dynamic and adaptive, which would inevitably reduce disease burden and improve survival and quality-of-life measures.
which are still in the process of being standardized, precluding comprehensive and collaborative analysis. [28] We recently identified cavitron ultrasonic surgical aspirator (CUSA) fluid as a rich, clinical source of glioma-derived EVs useful for biomarker discovery. [29] CUSA is a surgical tool designed to fragment and aspirate solid tumors of the CNS. [30] The system gently and precisely disaggregates and aspirates tumor tissue, while avoiding damage to normal surrounding brain tissue. [31] This procedure often results in up to 500 mL of irrigated fluid and blood, abundant with tumor tissue fragments and viable cells. CUSA tissue fragments are routinely captured for diagnostic histopathology, [32] while the cells are useful for molecular diagnostics, flow cytometry, animal xenograft studies, cell culture, and downstream in vitro studies. [31,32] The surgical resection of brain tumors also presents a unique and critical window in the search for biomarkers, as it is the only time at which the tumor microenvironment is directly accessible. Moreover, it represents a key time point in the patient's clinical management in that it is by far the most significant reduction of tumor mass, as well as a standardized point from which to measure time to recurrence, and thereby the effects of any postoperative treatment regimens on this. To confirm that CUSA fluid is a suitable source of glioma-derived EVs for biomarker discovery, we cultured cells isolated from a GBM CUSA washing in vitro and compared the proteome profiles of EVs released by these cells to those captured directly from CUSA fluid, as well as to our previously reported in vitro GBM EV proteome signature. [29] Using an optimized density-gradient ultracentrifugation EV enrichment protocol, we performed quantitative MS-based proteomics on CUSA-EVs captured from GBM and grade II-III glioma surgeries to identify biomarkers associated with more aggressive disease. Identification of putative biomarkers here could be traced to the peripheral circulation for www.advancedsciencenews.com www.proteomics-journal.com targeted liquid biopsies or support the development of GBMspecific EV enrichment strategies.

Surgical Specimens and Cohorts
CUSA washings were collected prospectively over a 24-month period (commencing October 2015). All patients provided research consent (Royal Prince Alfred Hospital Neuropathology Tumour and Tissue Bank, SLHD HREC protocol X14-0126) and specimens were processed and analyzed under approved University of Sydney HREC project 2012/1684. Diagnoses were confirmed histopathologically and included primary WHO grade IV GBM (n = 15) and grade II-III glioma (astrocytoma (n = 4) and oligodendroglioma (n = 3)). All GBM tumors were IDH-wild-type, while all grade II-III lesions harbored IDH mutations and the oligodendrogliomas were 1p/19q codeleted. A summary of the cohorts is provided in Table 1 (for more detailed information see Table 1a, Supporting Information).

EV Characterization by NanoSight Tracking Analysis & Transmission Electron Microscopy
The size distributions and concentrations of the CUSA-EV fractions were measured by nanoparticle tracking analysis software (NTA, version 3.0) in triplicate using the NanoSight LM10-HS (NanoSight Ltd, Amesbury, UK), configured with a tuned 532 nm laser and a digital camera system (CMOS Trigger Camera). EVs were diluted in sterile-filtered PBS (viscosity 1.09 cP) to ensure 20-100 particles were detected within the field of view of the standard CCD camera of the microscope. The NTA software captured triplicate 60 s video recordings of the EV, at 25 frames per second with default minimal expected particle size, minimum track length, blur setting, with the temperature of the laser unit controlled to 25˚C. The videos were analyzed by NTA3.0, which translates the Brownian motion and light scatter properties of each individual laser-illuminated particle into a size distribution (ranging from 10 to 1000 nm) and concentration www.advancedsciencenews.com www.proteomics-journal.com (particles per milliliter) while simultaneously calculating their diameter using statistical methods. [34] For transmission electron microscopy (TEM), EVs were resuspended in dH 2 O, loaded onto carbon-coated, 200 mesh Cu formvar grids (#GSCU200C; ProSciTech Pty Ltd, QLD, Australia), and fixed with 2.5% glutaraldehyde in 0.1 m phosphate buffer (pH7.4). Samples were negatively stained with 2% uranyl acetate for 2 min and dried for 3 h at RT. The samples visualized at 40 000× magnification on a Philips CM10 Biofilter TEM (FEI Company, OR, USA) equipped with an AMT camera system (Advanced Microscopy Techniques, Corp., MA, USA) at an acceleration voltage of 80 kV.

Preparation of EV Proteomes and LC-MS/MS Analysis
EV proteins were solubilized with 0.2% (w/w) Rapigest (Waters Corporation, Milford, MA, USA) in 0.05 m tetraethylammonium bicarbonate (TEAB) and incubated at 95°C for 5 min. To aid resuspension, the EVs were sonicated using a step-tip probe at 30% intensity for 15 s, placed on ice until cool, and sonicated again for a further 15 s. Proteins were reduced in 12 mm tris(2carboxyethyl)phosphine (TCEP) at 60°C for 30 min, then alkylated in 50 mm iodoacetamide in the dark for 30 min at RT. Sample pH was adjusted to 7.5 with 0.05 M TEAB, sequencing grade trypsin was added at a ratio of 1:30 (w/w) trypsin:protein, and samples were incubated overnight at 37°C. The following day, Rapigest was cleaved by acidification using 50% (v/v) TFA and removed by centrifugation (13 000 rpm, 10 min, 4°C). The supernatants were desalted by solid phase extraction using 1 cc HLB cartridges (Waters, MA, USA) and peptides were eluted in 70% ACN/0.1% formic acid (FA) (v/v). The peptides were quantified by a Qubit 2.0 Protein Assay and were dried by vacuum centrifugation. Peptides (0.6 μg) were analyzed using a Q-Exactive hybrid quadrupole-orbitrap mass spectrometer (Thermo Scientific). Peptide mixtures resuspended in 3% (v/v) ACN/0.1% (v/v) FA were separated by nano-LC using an Ultimate 3000 UHPLC and autosampler system (Dionex, Amsterdam, the Netherlands). RP mobile phase buffers were composed of A: 0.1% (v/v) FA (Thermo Scientific, Cat# 85178) and B: 80% (v/v) ACN (Thermo OPTIMA LC-MS grade, Cat# 34851-4), 0.1% (v/v) FA. Peptides were eluted using a linear gradient of 5% B to 42% B over 120 min with a constant flow rate of 250 nL min -1 . High voltage (2000 V) was applied to a low volume tee (Valco, Houston, TX, USA) and the column tip positioned 0.5 cm from the heated capillary (T = 275°C) of the mass spectrometer. Positive ions were generated by electrospray and the Orbitrap operated in data-dependent acquisition mode. A survey scan of 350-1550 m/z was acquired. Up to ten of the most abundant ions (>5000 counts) with charge states ࣙ+2 were selected and fragmented by CID (activation time of 10 ms). Mass-to-charge ratios selected for MS/MS were dynamically excluded for 20 s. Prior to loading the samples, an LC-MS/MS standard consisting of 30 fmol pre-digested BSA (GeneSearch, Cat# P8108S, 500 pmol) was injected to test the performance and dynamic range of the instrument. Due to limited sample yields of enriched EVs, single injections per specimen were analyzed by LC-MS/MS. Three repeat injections of specimen GBM-2 were performed sequentially and then over different time points (0, 60, and 120 h; total of five injections) to ensure repeatability and reproducibility of our MS quantitation approach. Normalized total spectra were plotted as scatter plots for all replicate pairs and the Pearson's correlation coefficient (r 2 ) was calculated in SPSS Statistics (IBM vr.22). Stable, reproducible, and repeatable MS/MS quantitation was observed between injections ( Figure 1, Supporting Information; r 2 > 0.987 for sequential injections, r 2 > 0.971 for injections over 120 h). Mascot .dat files were imported into Scaffold (version 4.8.7, Proteome Software Inc, OR, USA) and analyzed against the Swis-sProt Database (SwissProt 2018 05) with X! Tandem (The GPM, thegpm.org; version CYCLONE (2010.12.01.1). Mascot and X! Tandem were searched with a fragment ion mass tolerance of 0.10 Da and a parent ion tolerance of 4.0 ppm. Oxidation of methionine and carbamidomethylation of cysteine were specified in X! Tandem as variable modifications. Peptide and protein identifications were validated in scaffold with the peptide [35] and protein [36] prophet algorithms, respectively. Peptide identifications were accepted if they could be established at greater than 95.0% probability with scaffold delta-mass correction while protein identifications were accepted if they could be established at greater than 99.0% probability with at least two peptides identified at 95.0% or greater. False discovery rate (FDR) thresholds were calculated by searching the data against a decoy database. Proteins were grouped with protein cluster analysis. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. [37] Protein ambiguity was removed by manually deselecting all peptides shared by two or more proteins.

Data Processing, Protein Identification, and Quantitation and Statistical Analysis
Differential protein abundances were determined using Normalized Spectral Abundance Factors (NSAFs) and Student ttests with Benjamini-Hochberg correction for multiple comparisons. Reported fold changes are relative to GII-III. For fold changes <1, the negative inverse of the ratio was calculated to better visualize the symmetry of protein abundance changes. Significantly changing proteins were divided into GBM-HIGH or GBM-LOW lists depending on direction of fold change (positive or negative), where GBM-HIGH and GBM-LOW proteins were confidently identified in at least 12/15 GBM patients and five of seven GII-III patients, respectively. Proteins and their functions were annotated using Functional Enrichment Analysis Tool (FunRich; v3.1.3) [38] and Ingenuity Pathway Hematoxylin and eosin stains of (b-1) tumor tissue fragment and (b-2) cells recovered from the same CUSA washing corroborates a GBM grade IV diagnosis. b-3) Cells recovered from the CUSA washing were expanded in vitro. Transmission electron microscopy allowed visualization of vesicles within "exosome" size distributions in (c-1). CUSA fluid EV preparations and c-2) EVs isolated from the culture supernatants of expanded CUSA cells. d) NTA measured modal EV population sizes of 85-100 nm. e) Venn diagram shows overlap between LC-MS/MS analysis of EVs isolated from expanded CUSA cells, directly from CUSA fluid and previously reported GBM EV signature proteins. [29] f) Funrich annotations of proteins profiled in EVs isolated from CUSA fluid and cultured cells shows similarly enriched biological processes.
Analysis (Qiagen Bioinformatics, USA). For the sake of readability, the gene names corresponding to proteins of interest were used. For gene and transcripts expressions, gene names are italicized.

Data and Software Availability
MS/MS data captured from CUSA-EV have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset accession number: PXD010247 (https://doi.org/10.6019/PXD010247).

In Silico Analysis
Relative gene expressions and DNA copy numbers for all eight Tcomplex protein 1 ring complex subunits were analyzed in silico using TCGA gene expression (Human Genome U133A Array) and DNA copy number data (Affymetrix SNP 6.0 platform) [39] via the Oncomine platform (www.oncomine.org; Compendia Biosciences, MI, USA) [40] and cBio Cancer Genomics Portal (http://www.cbioportal.org/). The cBio Cancer Genomics Portal was used to observe associations between EGFR and CCT6A amplification. A gene-based Kaplan-Meier analysis was performed to test the association between CCT6A expression and overall survival using a log-rank test using continuous TCGA U133A Array data in SPSS Statistics (IBM vr.22). Kaplan-Meier plot was generated using averaged CCT6A levels dichotomized to high and low values based on median levels of the entire GBM cohort (n = 544).

CUSA Fluid, A Rich Source of Brain Tumor Derived-EVs for Biomarker Discovery
CUSA washings are a mixture of irrigation fluid and blood, rich with tumor tissue fragments that are used routinely for diagnostic histopathology [31] (see Figure 1a. for depiction of CUSA contents; Figure 1b1 for hematoxylin and eosin of tumor tissue fragment confirming a GBM diagnosis). CUSA washings are also an abundant source of live tumor cells (>95% viability; Figure 1b2.) that are useful for primary cell expansion. We established growth of cells rescued from CUSA fluid extracted during a GBM surgery in vitro (Figure 1b3.). EVs isolated from cultured CUSA cells had similar size distributions and morphologies to EVs captured from CUSA fluid (Figure 1c and d.). We then compared proteomes between EVs isolated from cultured cells with those captured directly from the CUSA fluid and observed 329 proteins identified in both preparations, 63% of total proteins identified in CUSA-cell EVs (Figure 1e). We observed a 79% (114) overlap between CUSA cell EV proteins and our previously reported in vitro GBM EV proteomic signature, which is composed of 145 proteins shared by EVs derived from six established GBM cell lines. [29] Funrich annotated the biological processes associated with identified proteins that were, on the whole, similarly enriched in CUSA fluid and CUSA cell EV preparations (Figure 1f). EV proteomes from GBM (n = 15) and GII-III (n = 7) were sequenced by LC-MS/MS and compared. Overall, 2967 proteins were identified at 99% confidence by at least two peptides (Table 2a, Supporting Information). Of these, 1480 proteins were confidently sequenced in EVs isolated from at least 12 GBM or at least five GII-III patients, with 885 proteins common to all patients (Figure 2a; Table 2a, Supporting Information). Significant overlap was also observed with "extracellular vesicle," "exosome," or "microvesicle" proteins recorded in Vesiclepedia (Figure 2a). Proteins identified in the GBM and GII-III specimens Figure 3. Significant, differentially abundant proteins in EVs isolated from GBM and GII-III neurosurgeries. a) Volcano plot of significant proteins (Benjamini-Hochberg CCT6A corrected p-value < 0.00496, significance threshold, green) with increased (red; GBM-HIGH) and decreased abundances (blue; GBM-LOW) in GBM relative to GII-III EVs. Proteins expressed only in GBM or GII-III EVs are listed by gene names in the red and blue boxes, respectively. b) Functional associations of significantly increased (red) and decreased (yellow) EV proteins in GBM specimens annotated by Funrich. c) Ingenuity pathway analysis revealed canonical pathways significantly associated with changing proteins, where positive (>2) and negative z-scores (←2) predict pathway activation (orange) and inhibition (blue) in GBM relative to GII-III specimens, respectively.

Characterization of Glioma-Derived EVs from Neurosurgical Aspirates
were similarly annotated by their sites of expression and involvement in biological processes and pathways, despite originating from different glioma subtypes (Figure 2b and c). GBM and GII-III EV proteins not included in Vesiclepedia were mapped to mitochondrial, lysosomal, endoplasmic reticulum, and membrane compartments with biological relevance in metabolism, energy pathways. and transport (Figure 3, Supporting Information).

T-Complex Protein 1 Ring Complex (TRiC) Subunits are Significantly Higher in Glioblastoma Tumors
Functional enrichment comparison of GBM-HIGH and GBM-LOW protein sets showed a large preponderance of proteins associated with protein metabolism in the GBM-HIGH group ( Figure. 3b). Specifically, these proteins included the T-Complex Protein 1 Ring Complex (TRiC) subunits. All eight subunits were sequenced by MS/MS and were more abundant in GBM EVs compared to GII-III. Six TRiC subunits, specifically TCP1, CCT2, CCT5, CCT6A, and CCT7 were significant at p < 0.00496, while CCT4 and CCT8 did not meet the significance threshold when adjusted for multiple testing (p < 0.05; Figure 4a). www.advancedsciencenews.com www.proteomics-journal.com Table 2. Potential EV marker proteins that discriminate GBM (IDH-wild-type) and Glioma (grade II-III; IDH-mutant) tumors.

Protein Name
Gene

GII-III (n)
Documented/potential roles in glioma
To determine whether these differences are the result of EV selective packaging or if they in fact reflect molecular changes associated with GBM pathology, gene expression levels, and DNA copy numbers of all eight subunits were interrogated in silico using TCGA data. [39] CCT2, CCT3, CCT5, CCT6A, and CCT7 gene expressions were significantly higher in GBMs relative to normal brain tissue (Figure 4b). All subunits with the exception of TCP1 showed significant copy number increases in GBM tumors relative to normal brain, and CCT2, CCT3, CCT5, CCT6A, and CCT8 were significantly increased relative to grade II-III astrocytomas (Figure 4c). Of the subunits, CCT6A showed the greatest induction of gene expression change in GBM specimens (2.97-fold, p = 3.2 × 10 −12 ) and displayed highly significant DNA copy number increases relative to healthy control brain (p = 1.01 × 10 −73 ) and grade II-III astrocytomas (p = 8.21 × 10 −31 ). Indeed, in GBM patients, CCT6A showed a higher rate of DNA copy number gain compared to the other subunits [ frequency of 10% or 57 GBM patients compared to 0-2.4% for the other TRiC subunits; n = 573; TCGA, 2013; Figure 4, Supporting Information).

CCT6A is an Independent Prognostic Marker and is Co-Amplified with EGFR
Using TCGA cohort survival data in univariate analysis, high CCT6A expression levels were associated with reduced patient overall survival (HR = 3.21; p = 0.006; Figure 4d). As CCT6A is located at the 7p11.2 locus, which also encodes EGFR, we interrogated the TCGA data further to identify any relationship between EGFR amplification (a common gene alteration in GBM) with increases in CCT6A copy number. Of the 57 GBM cases with CCT6A amplification, 54 also displayed EGFR amplification (odds ratio > 3, adjusted p-value < 0.001), indicating a strong tendency toward co-occurrence of EGFR and CCT6A amplification (Figure 4e).

CCT6A Tissue Distribution and EGFR Co-Expression
Immunohistochemistry of CCT6A expression in GBM and astrocytoma tumor tissues corroborated the proteomics measurements of CCT6A in CUSA-EVs. GBM tumor specimens (Figure 5a-1, a-2, b-1, b-2, c-1, and c-2) showed strong cytoplasmic CCT6A immunoreactivity with relatively weaker CCT6A expression observed in grade II astrocytoma specimens ( Figure  5e and f). GBM specimens immuno-positive for CCT6A also showed strong EGFR tissue expression (Figure 5a-3, b-3, and c-3; only Figure 5b was EGFR-amplified; Table 1B, Supporting Information). An additional GBM specimen with known EGFR immuno-negativity was tested and showed an absence of CCT6A staining (Figure 5d), indicating a potential link between EGFR and CCT6A protein expressions. Of note, EGFR was detected in GBM and GII-III EVs but levels were not significantly different between the cohorts.

Discussion
We have demonstrated that neurosurgical CUSA washings are a viable source of glioma-derived EVs for discovery proteomics. Cells isolated from CUSA fluid were readily expanded and the proteome of EVs released by cultured cells was similar to an in vitro GBM EV protein signature. [29] EVs from the expanded CUSA cells shared 329 proteins with EVs isolated directly from CUSA fluid from the same patient. We employed density gradient ultracentrifugation prior to our discovery proteomics analysis of CUSA-EVs from GBM and GII-III patients. This allowed EVs to settle at a higher density compared to highly abundant serum-related proteins present in the CUSA fluid. [41] Our methods, detailed here, allowed the reliable isolation and enrichment of sub-populations of EVs from heterogeneous clinical samples and ensured that EV preparations were better suited for discovery proteomics. The EV research field is rapidly expanding and the existence of diverse EV subtypes with distinct biophysical and biochemical properties is becoming evident. It is important to note that the methods employed here do not separate subtypes of small EVs or other structurally and functionally distinct nonmembranous particles, e.g., exomeres. [42,43] While comparative proteomic analysis of GBM and GII-III allowed the identification of proteins associated with the highly lethal and aggressive GBM phenotype, sample heterogeneity Significantly higher expression to normal brain ( * ) and grade II-III astrocytoma ( # ) for gene expression and DNA copy number is denoted by */ # p-value < 0.05, **/ ## p < 5.0 × 10 −5 , and ***/ ### p < 5.0 × 10 −8 . Gene expression levels of CCT2 (1.55-fold, p = 6.3 × 10 −7 ), CCT3 (1.36-fold, p = 8.12 × 10 −8 ), CCT5 (1.16-fold, p = 5.1 × 10 −6 ), CCT6A (2.97-fold, p = 3.2 × 10 −12 ), and CCT7 (1.26-fold, p = 8.76 × 10 −8 ) were significantly higher relative to controls. DNA copy numbers were significantly higher in GBM relative to controls (CCT2, p = 2.3 × 10 −5 ; CCT3, p = 1.1 × 10 -45 ; CCT4, p = 6.4 × 10 −17 ; CCT5, p = 1.9 × 10 −9 ; CCT6A, p = 1.01 × 10 −73 ; CCT7, p = 8.7 × 10 −16 ; CCT8, p = 1.1 × 10 −7 ) and grade II-III astrocytomas (CCT2, p = 0.003; CCT3, p = 0.001; CCT5, p = 1.1 × 10 −4 ; CCT6A, p = 8.2 × 10 −31 ; CCT8, p = 0.016). d) High CCT6A expression is significantly associated with reduced GBM patient overall survival (p = 0.006). Gene expression levels were dichotomized to high (green) and low (blue) values based on median population expression levels. e) Amplification of EGFR and CCT6A in GBM tumors shown as OncoPrints (cBioPortal for Cancer Genomics; www.cbioportal.org), where each bar represents a tumor that was found to contain a DNA alteration (amplification, deletion, mutation, as indicated) in EGFR and CCT6A; note that aligned bars represent the same tumor. Data are from TCGA, Cell 2013 resource, obtained through cBioPortal. In 54 cases, the genes were co-amplified; odds ratio > 3, adjusted p < 0.001. EGFR, but not CCT6A, was frequently mutated in addition to being amplified, as indicated by the green features. must be acknowledged. The GII-III cohort was composed of astrocytoma and oligodendroglioma tumors. While both subtypes harbor IDH-mutations and can progress to grade IV disease these glioma tumors are distinct entities [44] and by grouping them together, results cannot be related to specific pathological processes. Nevertheless, the work presented here provides insight into EVs isolated from neurosurgical aspirates, as well as the identification of proteins associated with a particularly deadly phenotype of what are highly variable tumors. As well as putative GBM EV invasion markers ACTR3, ANXA1, ITGB1, and PSMD2, [29] several key proteins previously linked to glioma aggressiveness, invasion, and/or poor prognosis were exclusively expressed in EVs captured during GBM surgeries, including S100A10, [45] S100A11, [46] CHI3L1, [47] CLIC1, [48,49] CAPG, [50] and MCT4 [51,52] (Table 2). Similarly, CBSL was reported as a novel prognostic marker for oligodendrogliomas, significantly upregulated in IDH-mutant relative to IDH-wild-type gliomas [53] and was expressed only in GII-III (IDH-mutant) EVs here.
Exosomal cargo reported in the literature is extremely diverse and it is, therefore, difficult to comprehensively attribute the presence of putative contaminants to bona fide contamination or simply a preponderance of cargo annotated to other principal sub-cellular localizations. The GBM-LOW proteins (significantly increased in GII-III) included several mitochondrial proteins that could indicate contaminating intracellular compartments in the GII-III specimens or simply reflect the energetic state www.advancedsciencenews.com www.proteomics-journal.com of GII-III tumors (Figure 3b and c). To complicate this further, extrusion and intercellular transfer of functional mitochondria occurs in vivo under various conditions, including cancer progression. [54] Indeed, GBM cells and normal astrocytes have been shown to release mitochondrial DNA (mtDNA) in EVs, [55] and astrocytes can transfer mitochondria to neurons, supporting neuronal viability after ischemic stroke. [56] The consistent presence of canonical exosomal markers, however, indicates a presumptive exosome/EV source for the proteins sequenced here, with a caveat for potential co-enrichment of intracellular proteins. Moreover, it is feasible that the significant overrepresentation of proteins involved in energy pathways and dysfunctional oxidative phosphorylation machinery in GII-III EVs is consequent to their IDH mutations. IDH mutant gliomas are known to harbor mitochondrial structural abnormalities, genomic mutations in mtDNA, and altered energy metabolism (Warburg effect). [57] Discovery analyses including well-defined cohorts of glioma subtypes with sufficient n, will allow the identification of biomarkers specific to glioma grade and phenotype.

TRiC is a Key Molecular Chaperone with an Emerging Role in Oncogenesis
Eight CCT subunits aggregate in a double torus structure to form the TRiC, which is a key eukaryotic chaperonin. [58] As a molecular chaperone, TRiC assists the folding of newly formed proteins, and in particular those rich in β-sheets, [59,60] which are often implicated in neurodegenerative pathologies. [61] This is accomplished in large part through the prevention of hydrophobic aggregation of nascent proteins by transient binding to exposed hydrophobic amino acid residues. [62] From there, ATPdependent conformational changes and subsequent dissociation with the nascent protein allow for proper folding. [63] Notably, a lack of chaperone activity increases the prevalence of misfolded proteins, as well as heterogeneous protein aggregates, which have a toxic effect on the cell. [62] Recent work has suggested that the degree of cytotoxicity is related to the degree of exposure of hydrophobic regions in extracellular protein aggregates, [64] although it remains to be seen whether this is relevant to intracellular aggregates as well. While TRiC appears to interact with 7-10% of proteins, [59,60] defects in its function are associated with increased toxicity [65] and a number of CNS pathologies. [66] The role of TRiC in cancer has recently gained more attention [60,67] as its interactions with a number of key oncogenic molecules have been elucidated. [68] Notably, it has been shown to contribute to the correct folding of p53, [69] where a reduction in TRiC activity in p53-mutant cells impaired cell invasion. Perhaps most interesting is TRiC's involvement in the development of the cytoskeletal architecture of the cell, including important roles in β-actin and αand β-tubulin formation and function. [70] Recent work has implicated disruption of the cytoskeleton in the efficacy of CT20, a peptide demonstrating selective cancer-cell toxicity. [71] Importantly, TRiC has been identified as the target of CT20; TRiC overexpression in breast cancer cell lines is required for increased susceptibility to CT20. [72] www.advancedsciencenews.com www.proteomics-journal.com Observed increases in the protein levels of all CCT subunits here (p < 0.05; six subunits p < 0.00496) suggests that CCT subunit stoichiometry is maintained, which is consistent with the notion that TRiC chaperonin activity is broadly activated in GBM tumors. Of further note is the number of significant, differentially abundant proteins with demonstrated associations with TRiC. These include the negative survival markers ANXA1, ACTR3, and PSMD2, all previously reported EV proteins that were significantly associated with GBM invasiveness. [29] Other significant proteins that are part of the TRiC interactome include GBM-HIGH proteins, ANXA2, EEF1G, EEF2, ICAM1, GNAS, GNG12, HSP90AA1, HSP90AB1, HSPA8, LGALS3, PSMC6, and RU-VBL1; and GBM-LOW proteins DNA-PK, PDHA1, OCIAD, SUN2, and XRCC6. [73] Interestingly, the predicted upstream regulators, MYC and KDM5 also bind TRiC subunits [74] and HSF1 and HSF2 transcriptionally activate all CCT subunits by binding to their heat shock element sequences. [75] Taken together, these observations portray a convincing role for TRiC in GBM pathology, both through relief of stress due to misfolded protein accumulation in cytoplasm, as well as the ability to impart a greater production of key cytoskeletal components, critical for migration and invasion in cancer cells. [76] Further studies, including the delineation of posttranslational modifications of TRiC proteins may help to resolve changes more functionally relevant to GBM progression. The presence of TRiC in GBM EVs further indicates the potential for tumor cells to confer their highly invasive phenotype, either in part or in whole, to neighboring cells. Moreover, they may be useful as biomarkers for the diagnosis of GBM or for the detection of glioma progression. Measurement of TRiC proteins may also indicate suitable candidates for CT20 therapy and should be further assessed by targeted assays using EVs from peripheral blood.

EV-Associated CCT6A as a Glioblastoma Biomarker
A growing body of research shows that individual CCT subunit genes are subject to copy number gains across many different cancers [77,78] and when alterations are present, changes in CCT subunits are largely mutually exclusive, [77] in line with observations here (Figure 4, Supporting Information). Yet, if subunit stoichiometry is required to generate functional TRiC, an increase in only one subunit would be futile. [77] The amplification, overexpression, and/or mutation of EGFR is a key genetic alteration in around 60% of GBM patients [79] and is associated with poor survival outcomes. [80] CCT6A and EGFR coamplification has been shown in non-small cell lung cancer [77] and low survival rates linked to high CCT6A levels in these patients. [81] As a proto-oncogene product, EGFR is an obvious driver of 7p11.2 amplification. CCT6A copy number and expression increases observed here in silico, might simply be a consequence of EGFR amplifications on flanking genomic sequences, as implied by the significant co-occurrence rate (Figure 4e). The converse may also be true, where genetic alterations leading to reduced EGFR expression might decrease the transcription and subsequent translation of neighboring CCT6A, as was observed immunohistochemically here (Figure 5d). As a tyrosine kinase receptor, EGFR is a prime anticancer drug target and CCT6A levels in EVs might allow selection of suitable patients for therapies targeting EGFR. There is, however, some evidence of functional differences among CCT subunits and for distinct physiological properties of the unassembled subunits, especially for CCT6A. [82] CCT6A was reported to be an essential promoter of the TGF-β-induced aggression and metastasis of lung cancer cells and a negative regulator of SMAD2. [83] Similarly, silencing of CCT6A decreased GBM cell invasion and impaired epithelialmesenchymal transition in vitro. [84] However, whether these observations were simply the result of reduced TRiC function and ability to fold important cytoskeletal elements was not tested.

Concluding Remarks
We have isolated clinical, glioma-derived EVs directly from the tumor microenvironment and optimized enrichment methods for biomarker discovery. EV characterization and subsequent comparative proteomic analysis identified 298 significantly changing proteins between the GBM and GII-III cohorts. These included previously identified putative EV biomarkers for GBM invasiveness (ANXA1, ITGB1, ACTR3, and PSMD2) among many other interesting targets. Levels of all eight subunits of the key molecular chaperone, T-complex protein 1 Ring complex (TRiC), were higher in GBM-EVs, including CCT2, CCT3, CCT5, CCT6A, CCT7, and TCP1 (p < 0.00496). Analogous increases in TRiC transcript levels and DNA copy numbers were also detected, where CCT6A showed the greatest induction of expression and amplification in GBM as well as a significant negative association with overall survival (p = 0.006). CCT6A is co-localized with EGFR at 7p11.2, with a strong tendency for coamplification (p < 0.001). Immunohistochemistry corroborated the CCT6A proteomics measurements and further indicated a potential link between EGFR and CCT6A tissue expressions. Further studies are needed to better understand this molecular interplay and may reveal EV-associated CCT6A as a proxy for EGFR testing. Putative EV-biomarkers described here will be further assessed in EVs captured from peripheral blood.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.