Fibroblast activation protein (FAP) is upregulated in most cancers and represents an emerging theranostic target in oncology. In glioblastoma (GBM), FAP is predominantly expressed in mesenchymal pericyte-like cells, which contribute to GBM progression by promoting angiogenesis and glioma cell proliferation by soluble mediators. In this work, we show another possible mechanism by which these cells contribute to gliomagenesis. We observed that FAP+ pericyte-like cells residing in the perivascular niche in GBM are associated with elevated levels of the fibrillar extracellular matrix (ECM) proteins collagen I and fibronectin. Using sequential window acquisition of all theoretical mass spectra (SWATH-MS) analysis, we confirmed that the highest levels of collagen I and fibronectin are produced by FAP+ pericyte-like cells. In addition, we observed different expression pattern of basal membrane proteins between FAP+ pericyte-like cells and glioma cells.

FAP+ pericyte-like cells derived from fresh GBM tissues and glioma cell lines were cultured on gelatin coated plates in their corresponding media (FAP+ pericyte-like cells: pericyte medium supplemented with 2% FBS, pericyte growth supplement and 100 units/mL penicillin G, and 100 µg/mL streptomycin; glioma cells: DMEM supplemented with 10% fetal calf serum). Media were supplemented with 50 μg/mL of ascorbic acid and were changed every other day. After 8 days, 3D ECMs were obtained by decellularization of plates with alkaline detergent treatment (0.5% Triton X-100, 20 mM NH4OH in PBS for 5 min at 37°C). Decellularized matrices were incubated with preheated lysis buffer (1% SDS in 50mM Tris HCl, 95°C) and harvested with cell scraper. SWATH-MS analysis was performed to elucidate differences in expression of ECM proteins between FAP+ pericyte-like cells and glioma cells.

ECM samples were sonicated (10 min, 40 kHz) and mixed with the washing buffer (8 M urea and 5 mM EDTA in 50 mM ammonium bicarbonate – AmBic). Microcon 30K centrifugal ultrafiltration units were used for filter aided sample preparation (FASP). The SDS-containing buffer was exchanged with the washing buffer, proteins were reduced by 10 mM tris(2-carboxyethyl) phosphine (30 min, 32°C) and alkylated by 40 mM iodoacetamide (35 min, 25°C, in the dark). Protein digestion was performed in 50 mM AmBic with 0.02% (w/v) ProteaseMAX and 1 µg of LysC protease (37°C, two hours), followed by addition of 1 µg of trypsin (37°C, overnight). Peptides were eluted by centrifugation, acidified with 1% formic acid (FA) and desalted using C18 MicroSpin columns. Desalted peptides were dried in vacuum centrifuge (45°C) and resuspended in a loading buffer comprised of 2% acetonitrile (ACN) in 0.5% FA. Peptide concentrations were determined at λ = 280 nm, and retention time normalization peptides (iRT, Biognosys) were added at iRT : sample ratio 1:20 to all samples. One µg of each peptide sample was separated in a trap-elute mode, using the Eksigent nano-LC 425 (Sciex) on-line connected to 5600+ TripleTOF (Sciex). Peptides were loaded on Acclaim PepMap 100 C18 trap column (5 μm, 0.1 × 20 mm; Thermo Fisher Scientific) for 10 minutes at 2 μL/min and then separated by fused-silica column (25 cm length, 75 μm inner diameter) packed in-house with ProntoSIL 120 Å 3 μm C18 AQ beads (Bischoff Analysentechnik GmbH). Separation was performed by linear gradient of ACN with a flow rate of 200 nL/min (5–35% ACN in 0.1% FA over 120 min, followed by 35–50% ACN in 0.1% FA over 10 min). For data-dependent acquisition (DDA) approach, 40 positively charged precursors with the highest intensity in MS1 (mass range 400-1250 Da, accumulation time 300 ms) were fragmented and analyzed in MS2 in high sensitivity mode (mass range 170-2000 Da, accumulation time 100 ms). The cycle time was 4.35 s, precursor exclusion time was 13 s. SWATH-MS method with 30 variable windows was employed for protein quantification. Variable windows were calculated by a SWATH Variable window calculator (Sciex) based on data from DDA measurement of ECM sample. SWATH-MS was acquired in the same mass range as DDA measurements. The accumulation time was 150 ms and 100 ms for MS1 and MS2, respectively, and cycle time was 3.2 s. The DDA data were analyzed in Mascot Distiller 2.7.1 and Mascot Server 2.6.2 (both Matrix Science Ltd.) for protein identification and sample-specific spectral library preparation using human reference proteome (UniProt, 20,607 proteins, one protein sequence per gene) with common protein contaminants. Mascot search mass tolerances were set to 15 ppm and 20 ppm for MS1 and MS2, respectively. Protein modifications were set to carbamidomethylation of cysteine (fixed), and protein N-terminal acetylation and oxidation of methionine (both variable). Enzyme specificity was set to always cleavage after lysine, and cleavage after arginine, unless proline preceeded, with two missed cleavages allowed. The SWATH-MS data were analyzed in Skyline-daily (version 22.2.1.351) using sample-specific spectral library created in Skyline from 16 DDA measurements searched by Mascot (Mascot expectation score threshold 0.01). MS2 filtering of data was performed with resolving power 25,000 in high-selectivity extraction mode. The iRT peptides were used to align retention times between samples and retention time filtering was set to select only peaks within ±5 min of predicted retention time. Duplicated and repeated peptides were removed and only proteins with ≥ two peptides (≥ three transitions per peptide) were further processed. The peptide decoys were generated by a shuffle sequence method to train the mProphet for automatic peak picking.