Study design
This investigation was performed after approval by a local Human Investigations Committee. This study was approved by the “Comité de Protection des Personnes” (ethical committee) and the “Commission Nationale de l’Informatique et des Libertés” (CNIL) according to French regulations, and all methods were performed in accordance with the ethical principles as contained in the Declaration of Helsinki. Written informed consent was obtained from all subjects and/or their legal guardian(s). https://clinicaltrials.gov/ registered number: NCT00773656 (Fig. 1). Following an over-irradiation during their treatment against a prostate adenocarcinoma, patients developed a chronic radiation enteritis RB with distinct severity degrees. Cohorts from Epinal hospital (France) were defined following the discovery of a number of technical problems that led to the over-irradiation of 208 patients: patients of the cohort 1 were subjected to an overdose of radiation from 20 to 30%; cohort 2 patients received an overdose of 8 to 10%; and cohort 3 patients received a radiation overdose of 3 to 7.1%. According to their RB grade, determined by CTCAE method, at the time of blood sampling23, patients were divided into 4 severity grades (0, 1, 2 and 3/4) which included 84, 91, 25 and 8 patients respectively. The CTCAE classification does not include grade 0 but in our study, the grade 0 patients included all patients who received an overdose of radiation but did not develop any symptoms at the time of blood sampling. Blood samples were collected in 0.129 M sodium citrate tube and all patients provided informed consent. Several comorbidity factors that may influence the level of circulating vesicles, and other characteristics of the study participants are shown in Table 1.
MV isolation, labeling and flow cytometry analysis
Platelet-free plasma was obtained after centrifugation of blood sample at 450g for 15 min at room temperature to remove cell debris, followed by another centrifugation at 13,500g for 5 min, 4°C. The supernatant was distributed into 1.5 mL tubes and stored at -80°C. MVs were isolated from those samples by centrifugation at 20,600g for 45 min, 4°C.
MVs were labeled with Annexin V–FITC (IgG1), CD14–PE (clone RMO52), CD31–PE (1F11), CD41–PC5 (P2), CD3–PE (UC HT1) and CD235a–PC7 (KC16) and their corresponding isotype controls (Beckman Coulter, Villepinte, France). Incubation was carried out for 30 min in the dark at room temperature and MVs were resuspended in Annexin V Binding Buffer 1X (Beckman Coulter). To determine the absolute number of MVs per microliter in samples, 30μL of counting beads with an established concentration (Flow CountTM Fluorospheres, Beckman Coulter) were added to each sample.
Analyses were performed on a Cytomics FC500 flow cytometer (Beckman Coulter). MV detection was performed as previously described24. The cytometer was calibrated using a mix of fluorescent beads (Biocytex, Marseille, France) of different diameters to determine a detection window. The 3-μm beads indicated where the Flowcount beads appeared while the 0.5 and 0.9 μm beads determined the lower and upper detection limits of MVs, respectively. Only events included within this gate were further analyzed.
Exosomes analysis
Size distribution and concentration of exosomes were determined using a NanoSight NS300 instrument (Malvern Panalytical, Malvern, UK). Vesicles in platelet free plasma were visualized and tracked by laser light scattering after a 1:100 dilution. Samples were analyzed with NTA 2.3 software (Malvern Panalytical).
Statistical analysis
EV counts were expressed as median and range (25th percentile and 75th percentile). Graphs represent the distribution of the populations with median (horizontal bar), 25th and 75th percentile (boxes), and 10th and 90th percentile (error bar). Analyses were performed by one-way ANOVA followed by Tukey’s multiple comparison test using GraphPad Prism software v.5.03 (GraphPad Software, San Diego, CA). When data did not show a Gaussian distribution, Kruskall-Wallis test was used followed by Dunn’s multiple comparison test. Results with a P-value<0.05 were considered significant.
Electronic microscopy
Pelleted MVs were fixed with a 2.5% glutaraldehyde solution for 1h at 4°C. After a centrifugation at 20,000g for 120 min at 4°C the solution was replaced by PBS. The pellet was solidified by addition of agarose before fixation with osmium tetroxide 2%.
Cell culture
The human U937 monocytic leukemia cell line (ATCC) was cultured with RPMI 1640 supplemented with FBS 10%, Hepes and Penicillin/Streptomycin (50 IU/ml and 50 μg/ml, respectively) (Life Technologies, Villebon sur Yvette, France). Primary human microvascular endothelial cells (HMVEC) and lymphatic endothelial cells (HLMVEC, Lonza Verviers, Belgium) were cultured with EBM-2 medium supplemented with hEGF (0,1%), hydrocortisone (0,04%); GA-1000 (0,1%), FBS (20%), VEGF (0,1%), hFGF-b (0,4%), R3-IGF-1 (0,1%) and ascorbic acid (0,1%) (Lonza). After thawing, all cells were cultured in 75-cm² flasks for amplification, then in 6-well plates by using Trypsin-EDTA (1X) 0.05% (Life Technologies). After reaching confluence state, cells were irradiated using a 137Cs source at 10Gy, 20Gy and 40Gy for all cell types.
MV-dependent functional assays
The MV-dependent TF activity was measured using a procoagulant activity assay, as described earlier.25 The MV-dependent plasmin generation assay was conducted using a chromogenic test, as previously described.26
MV proteomics analysis
Protein extraction, tandem mass tagging labeling, LC-MS/MS analysis, assignment of MS/MS spectra, and quantitative data analysis were performed as previously described, starting from 50 µg MV lysates.27 Statistical analysis of radiation induced protein changes was conducted using the LIMMA package of R framework and Benjamini-Hochberg correction for multiple comparisons. 28
Association between organ irradiation and EV secretion
The association between the organ dose distribution and MV counts, or with exosome enumeration, was investigated through a scalar on function regression.29,30 In this approach, the dose distribution is represented by the quantile function which achieves a “synchronization” between the irradiation profiles in order to exploit the patient-to-patient irradiation heterogeneity in the analysis.31
The general formulation of a scalar on function regression model is given by:
EV secretion = ∫ βDose(x) × QDose(x) dx + 𝐶𝑜𝑣𝑎𝑟𝑖𝑎𝑡𝑒𝑠
where QDose(x) is the quantile function predictor and βDose(x) is the functional parameter of the model providing an exposure weighting over the whole range of dose values. A positive (negative) value of the functional parameter βDose(x) can be interpreted as an increase (decrease) of EV secretion when the organ is exposed to the corresponding quantile doses. Note that additional confounding factors can be included in the model (age, diseases, concomitant treatments, etc). The functional parameter was estimated using penalized spline functions using the REFUND Package of R software.28
The dose distribution to the anal canal, anterior rectal wall and bladder was assessed through relative dose volume histograms (DVHs), in 0.03 Gy dose bins, constructed for each patient using treatment planning systems Eclipse and CadPlan from Varian Medical Systems (Palo Alto, CA).
Association between EV secretion and RB grade during the post blood sampling follow-up
The association between the EVs and a worst RB grade during the post blood sampling follow-up was investigated using a multivariate logistic regression (Table 2). In this analysis, an event consists in developing a more severe RB during the post blood sampling period compared to the RB grade scored at the time of blood sampling. The model was adjusted on age, prescribed radiation dose, the delay between the radiotherapy and the blood sampling as well as the toxicity grade at the time of blood sampling. Adverse treatment such as anticoagulants, anti-aggregates and anti-hypertensives, laser treatment or transfusion of packed cells were tested and model selection was performed using the Akaike information criteria.32
MV and exosome data were incorporated in the model both in a quantitative (using a log-transformation) and compositional manner. Based on the cellular origin of MVs obtained by flow cytometry analysis, the MV composition of each patient was summarized by a four-component vector, listing the proportion of MVs derived from platelets, endothelial cells, monocytes and others. These compositional data were then treated using the Aitchison geometry, as described.33 In the present study, we adopted an isometric log-ratio transformation for an isometric mapping from the simplex sample space to the real space.34