Proteomics analysis of serum small extracellular vesicles for the longitudinal study of a glioblastoma multiforme mouse model

Longitudinal analysis of disease models enables the molecular changes due to disease progression or therapeutic intervention to be better resolved. Approximately 75 µl of serum can be drawn from a mouse every 14 days. To date no methods have been reported that are able to analyze the proteome of small extracellular vesicles (sEV’s) from such low serum volumes. Here we report a method for the proteomics analysis of sEV's from 50 µl of serum. Two sEV isolation procedures were first compared; precipitation based purification (PPT) and size exclusion chromatography (SEC). The methodological comparison confirmed that SEC led to purer sEV’s both in terms of size and identified proteins. The procedure was then scaled down and the proteolytic digestion further optimized. The method was then applied to a longitudinal study of serum-sEV proteome changes in a glioblastoma multiforme (GBM) mouse model. Serum was collected at multiple time points, sEV’s isolated and their proteins analyzed. The protocol enabled 274 protein groups to be identified and quantified. The longitudinal analysis revealed 25 deregulated proteins in GBM serum sEV's including proteins previously shown to be associated with GBM progression and metastasis (Myh9, Tln-1, Angpt1, Thbs1).


Methods S1. Mice Tumor induction
The murine glioma GL261 cell line were grown as specified in Vannini et al., 2014 [1]. C57BL/6 mice were anesthetized with avertin (intraperitoneal injection of 2,2,2-tribromoethanol solution; 250 mg/kg body weight). The tumour was induced by stereotaxically guided injection of 40,000 GL261 cells (20,000 cells/1 μl Tris HCl solution) into the primary motor cortex (1.75 mm lateral and 0.5 mm anterior to bregma). Body temperature was constantly monitored with a rectal probe and maintained at 37.0°C with a thermostatic electric blanket during the surgery. An oxygen mask was placed in front of the animal's mouth to aid respiration. The GL261 cell solution was slowly delivered at a depth of 0.8-0.9 mm from the pial surface. To prevent dehydration, a subcutaneous injection of saline (0.9% NaCl, 1 ml) was delivered at the end of the procedure. the ExoCarta Top100 database [2]. Sample EV-1 contained the highest number of EV-proteins, which was more than double that identified from any other SEC sample and also contained the most unique protein identifications ( Figure S1). The total protein amount extracted from SEC sample EV-2 indicated serum protein contamination, 3.5 µg in comparison to 1.8 µg for EV-1 (Table S1). Serum proteins also dominated the late SEC eluate samples Prot-1 and Prot-2 ( Figure   S1). In view of the higher purity of the sEV's all subsequent SEC experiments were performed using the EV-1 protocol.

Methods S3. Proteomics sample preparation
The sEV lysate was obtained by on-filter buffer exchange and on-filter sEVs lysis. Protein digestion was performed using a modified SP3 protocol [3] , [4] , [5] that was further adapted in this work owing to the higher sample volume (sEV lysate collected from 3KDa filter was approximately 50 µl). The protein extract solution was mixed with trifluoroethanol (TFE) in a 1:1 ratio and 2 µl of carboxylate coated paramagnetic beads (100 mg/ml solution of 50% Speedbeads RT (30 sec on/off cycles). The total protein content in each sample was then quantified using 1 µl aliquots and a modified microBCA assay [3]. Enzymatic digestion was performed overnight (18h) at 37°C in 12 µl by adding Trypsin/Lys-C mixture (1:25 enzyme/protein). Digestion was optimized in this work by testing two different digestion conditions (

Methods S5. Data analysis -Linear mixed effect model (LME)
For the longitudinal study, identified protein groups were filtered by eliminating serum albumin, and immunoglobulins, whose intensities may depend on the serum preparation and SEC elution.
The data were then log2 transformed and normalized by median subtraction. An ANOVA test could not be used for the comparison between the three different time points, because the samples are not independent (longitudinal analyses explicitly concern the time dependence of the signals).

Results S1. Comparison of serum and plasma SEC-EV proteins
During the preparation of serum the fibrin clot incorporates a large number of blood proteins [8], and extracellular vesicles may become trapped within this network [9]. Differing amounts of fibrin during clotting could thus affect the number of EVs that can be isolated from serum. To assess if this source of variability had an effect on the analysis of EV proteins the SEC-EV isolation procedure was applied to 50 µl serum and 50 µl plasma samples from the same mice (n=3), the proteins were then extracted from the sEV's and digested using the two-step digestion procedure.
The number of identified proteins was much more variable for the plasma derived sEV, with 213±4 protein groups identified from serum-sEV and 206±88 from plasma-sEV, Figure S2A. Protein intensities were also more variable for the plasma-sEV. Figure S2B shows the result of principal component analysis (PCA) performed on the log2 transformed, median normalized data. The PCA score plot shows that the two sample types are separated by PC1, and that the serum sEV samples from different animals were clustered together more closely than the plasma sEV samples. The loading plot shows that three proteins were primarily responsible for the separation of plasma sEV's from serum sEV's, Figure S2C, and were all part of the fibrinogen family (Fga, Fgb, Fgg). Figure S1. Supplementary Tables   Table S1. LC-MS/MS proteomics analysis of the SEC elution fraction pools.