Autophagy inhibition reinforces stemness together with exit from dormancy of polydisperse glioblastoma stem cells

Down regulation of autophagy awakened dormant GB CSCs

In a previous work [32], we demonstrated the synergistic action of autophagy and the neurotrophin signaling pathway (i.e. TrkC) in GB cell survival, underscored by the exacerbated expression of these pathways in patient samples. In order to evaluate the involvement of autophagy and neurotrophin in CSCs maintenance and behavior and/or in the cellular ecology of glioblastoma, we constructed U87sh cell lines (U87shBeclin1, shATG5 and shTrkC) and checked the down-regulated expression of targeted proteins (Figure 1A–1D). Indeed, even if Beclin1 is described as being a determinant actor in autophagy initiation, we decided to check the data by silencing ATG5 a protein with a later onset of action in the process i.e. elongation of the isolation membrane.. Among the 4 shATG5 clones tested, the clone 165 led to a deeper silencing than the others (Figure 1D) and has been therefore chosen for further experiments. As shown in cell culture pictures (Figure 1A), the shBeclin1 cells were the most phenotypically modified cell line showing a higher propensity to develop spheres, a characteristic of immature cells. In comparison, the shATG5 cells appeared smaller than U87pLKO cells, without sphere formation.

These observations led us to analyze the classical CSCs markers’ expression, i.e. BMI1, nanog, oct4, sox2 transcription factors and the WNT pathway co-receptor LGR5 (Figure 2). First of all, we checked that hypoxic conditions led to expected increase of stemness in U87 pLKO cell line. As shown in Supplementary Figure 1, expression of LGR5 and BMI1 cancer stemness markers was significantly enhanced when cells were cultured under hypoxic conditions p<0.05. Using RT-qPCR analysis, it appeared that U87shATG5 modified cell line expressed more intensely stem cell markers in comparison with U87pLKO either in normoxia or hypoxia, i.e. 1% O₂ (Figure 2A). A similar enhancement was also observed when silencing another autophagy related gene, Beclin1 both in normoxia and hypoxia (Figure 2B). Western blot analysis of stemness markers, also exhibited an increased expression of BMI1 for both modified cell lines and of LGR5 for U87shATG5, whereas Oct4 expression failed to vary (Figure 2C).

To check and extend these results, we used another GB cell line i.e. LN18 (Figure 1A), in which autophagy was knocked down by the means of ATG5-targeting shRNA (Figure 1E). Similarly to U87 cell line, the clone 165 was more efficient for reduction of ATG5 expression and was chosen for further experiments. Concordant results at transcriptomic level were obtained, with significant enhancement of SOX2 expression in normoxic conditions and of SOX2, NANOG and POU5F1 in hypoxia (Figure 2D). This first set of data showed in two GB cell lines, that autophagy down regulation (by targeting two different genes) reinforces stemness phenotype.

In order to confirm these results, we underwent a complementary study where specific differentiation markers of astrocytes (GFAP and Doublecortin) were assessed by the flow cytometry analysis in parallel to Oct4 and the classical stemness marker, CD133 (Figure 3 and Supplementary Figure 2). Furthermore, in addition to genetic autophagy silencing (shBeclin1 and shATG5), we used chloroquine (CQ) -treated U87pLKO cells, a well-known inhibitor of the autophagic flux. The enhancement of CSCs markers expression, Oct4 and CD133 was evidenced in the three conditions where autophagy was down-regulated (Figure 3, upper panel). In parallel, expression of the differentiation markers GFAP and DCX was impaired, especially in genetically modified cell lines (Figure 3, lower panel). To conclude, in response to this multi-way inhibition of the autophagic process (shBeclin1, shATG5 or CQ treatment) in two GB cell lines (U87 and LN18), the population of CSCs seemed to be potentialized rather than reduced.

Considering that the expression of stemness markers reflects a steady state of CSC, we explored the functional properties of the U87shBeclin1 cell line. We chose this modified cell line since when considering the morphology, the U87shBeclin1 was the only one to generate spheres, a characteristic of immature cells (Figure 1A). We analyzed cell proliferation with BrdU incorporation, at 96 h under normoxic and hypoxic (1% O₂) conditions (Figure 4A) and at 144 h under normoxia only (Figure 4D). Both conditions increased cell proliferation of the U87shBeclin1 cell line, which reached a significant level under hypoxia at 96 h (**p<0.01 vs pLKO) and under normoxia at 144 h (***p<0.001). These results were confirmed by Ki67 staining in both shBeclin1 and shATG5 U87 modified cell lines (Figure 4B). This higher proliferation rate could not be attributed to an enhancement of MAP kinase or AKT pathways (Supplementary Figure 3). We also tested the clonogenic properties of U87shBeclin1 cells with CFU assay in agar matrix and evaluated colony numbers as well as colony areas, two parameters that characterize active CSCs. Down regulation of the autophagic process increased, at the same time, the number and the size of U87shBeclin1 colonies, in comparison with control (Figure 4C, ***p<0.001), hence corroborating proliferation rate enhancement. Since chemotherapeutic resistance is among the best known of the CSCs functional characteristics, we examined BrdU incorporation (cell proliferation) and apoptotic cell death after temozolomide (TMZ) treatment (48 h; 1.5 mM; Figure 4D, 4E). As expected, TMZ significantly decreased cell proliferation in both cell lines (control and U87shBeclin1). When we evaluated the apoptotic level by ELISA cell death after TMZ treatment, a significant increase of apoptosis was observed in U87pLKO (** p<0.01, Figure 4E) which was not the case of the U87shBeclin1 cell line. Indeed, this cell line exhibited a moderate apoptosis capability in whichever condition (***p<0.001).

As previously mentioned, our precedent work highlighted an exacerbated activation of the TrkC-NT3 axis in GB cell line and patients’ samples (32). That’s why, in complement to autophagy inhibition, the same analysis was done on U87shTrkC cells (Figure 1C). When the expression of the receptor was down regulated, we failed to observe neither variation in CSCs markers’ expression (Supplementary Figure 4A), nor in clonogenicity (Supplementary Figure 4C). Variations of proliferation (96 h, Supplementary Figure 4B and 144 h, Supplementary Figure 4D) and TMZ-induced apoptosis (Supplementary Figure 4D), seemed to be similar to the variations observed when autophagy was inhibited, although staying somewhat weaker.

Thus, since our initial goal was to affect CSC compartment, we chose to favor the autophagy inhibition approach in the subsequent study instead of knocking-down the TrkC signaling pathway. Indeed, inhibition of autophagy by the means of two different shRNAs showed an enhancement of the expression of CSCs markers and led to the forced awakening of CSCs, from the dormant state. So, we hypothesized that the different microstates of their CSCs were modified rather than a global modification of the overall population.

SdFFF GB cell sorting allows to reveal CSCs heterogeneity

In order to better characterize the CSCs population, we enriched the different fractions with the help of SdFFF. This tool allows cell separation depending on the combined action of a parabolic profile generated by the flowing mobile phase through the channel and a multigravitational external field (generated by rotation of the channel) applied perpendicularly (large red arrow on the depletion wall on Figure 5A) to the flow direction. Thus, larger and less dense cells generating more hydrodynamic lift forces within the mobile phase and not very sensitive to centrifugal force (blue cells on Figure 5A), will be eluted before smaller and denser cells (red cells, Figure 5A) generating less hydrodynamic forces and more sensitive to the multigravitational field. As for a number of other cell models [27, 29, 31], we found that under the elution conditions, U87 cells were eluted under the hyperlayer elution mode [28, 30] schematized in Figure 5A.

In each case, we observed the classical two-peak profile (Figure 5B); the first peak corresponds to the void peak (VP, t₀/t₀ = R_obs ≈ 1) containing unretained species (cell debris, proteins, cell aggregates…), and the second corresponds to the cell peak (CP, t₀/t_r = R_obs < 1). These broad cell peaks (Figure 5B) correspond to very polydisperse populations, reflecting the heterogeneity of these GB cell lines containing a cell continuum, from fully differentiated cells to cancer stem cells [28, 30, 35].

As expected, elution profiles of the U87MG and U87pLKO cell lines are similar and characterized by quite identical retention ratios, R_obs = 0.319 ± 0.010 and R_obs = 0.320 ± 0.012 respectively. An important difference in elution profile was observed with U87shBeclin1 cells whose R_obs = 0.240 ± 0.006, corresponded to an important shift quantified by ∆R_obs = R_{obs pLKO} - R_{obs shBeclin1} = + 0.079 ± 0.014 (Figure 5B). This right shift led to a better resolution of the cell peak, well separated from the void volume, and associated with a flattened signal. In each one of the three fractograms, and according to the hyperlayer elution mode, the cells collected in fraction 1 (F1) are larger than those collected in the last fraction (F3). Interestingly, we observed that the “s” parameter of U87pLKO cells was larger than their average radius "r", while this is not the case of U87shBeclin1 cells, suggesting an increase of their retention time (≈ 1 min), without an important decrease in cell diameter. This is particularly true for fraction F3, as F3 U87shbeclin1 cells had a similar size to those of F3 U87pLKO (Figure 5C), demonstrating here that autophagy inhibition led to an important increase in cell density. This is usually associated with an increase in stemness properties. Once again, in order to check the strength of this peak shift specifically observed with the U87shBeclin1, we subjected U87shATG5 and CQ-treated U87pLKO cell lines to SdFFF (Supplementary Figure 5). In all the cell lines in which autophagy was down regulated, similar results were observed, the fractograms exhibited right-shifted with roughly similar ∆R_obs (Supplementary Figure 5). The same experiments have been conducted with the U87shTrkC cell line, but they failed to detect such differences (Supplementary Figure 6). The narrowing and the lowering of the second peak suggest a homogenization of the cell population, including the CSCs pool, which is highlighted by a left shift of the peak (∆ R_obs = R_{obs pLKO} - R_{obs shTrKC} = - 0.034 ± 0.014) (Supplementary Figure 6).

SdFFF allows the isolation of a subset enriched in proliferative and clonogenic CSC

We wondered whether the biophysical changes revealed by SdFFF analysis could help us to distinguish among sub-cellular populations, especially CSCs. As expected, and as already published [28], the F3 U87pLKO cell fraction exhibited more CSCs markers as compared with cells of the F1 fraction (data not shown). In the cell line in which autophagy was down-regulated (U87shBeclin1), analysis of CSCs markers confirmed a weak enrichment of the F3 fraction, based on enhancement of LGR5 and POU5F1 expressions (*p<0.05 and **p<0.01 respectively, Figure 6A). Expressions of the transcription factors BMI1 and SOX2 in the F1 and F3 fractions seemed to be fairly similar.

Comparison of clonogenic assays performed with cells of the F1 and F3 fractions obtained from U87shBeclin1 and U87pLKO cells showed a significant CSCs enrichment of F3 (U87shBeclin1) as compared with F3 (U87pLKO) (Figure 6B). Indeed, we observed that cells from F3 (U87shBeclin1) displayed increased colony size and number as compared with cells from F3 (U87pLKO) (***p<0.001, Figure 6B). Finally, we checked the impact of TMZ on these cellular fractions using BrdU incorporation and nucleosome quantification (Figure 6C). TMZ treatment had a specific impact on F1 and F3 cell proliferation in the two cell lines (Figure 6C left panel). F3 cells from U87pLKO appeared to be less proliferative than F1 cells (***p<0.001), whereas U87shBeclin1 F3 cells seemed to proliferate more than F1 cells (**p<0.01). Using “Genomics of Drug Sensitivity in Cancer” [36] (Supplementary Figure 7), U87-MG drug sensitivity demonstrated that temozolomide is still the gold standard of chemotherapy against glioblastoma. Nevertheless, in our experimental conditions (1.5 mM TMZ), pLKO appeared to be more prone to TMZ-induced apoptosis as compared with the shBeclin1 cell line, whatever the fraction considered (Figure 6C, right panel). As expected, the U87shbeclin F3 fraction, enriched in CSCs cells, exhibited the weakest apoptotic rate, when treated with TMZ (***p<0.001, F3 pLKO vs F3 shBeclin1, both TMZ-treated). Such a result showed that the CSCs enriched fraction issued from the lineage defective in autophagy summarized the data obtained with the whole population: expression of stemness markers was amplified, and CSCs came out from dormancy, being more proliferative and clonogenic. In addition, as described in Colorectal Cancer [29], we can distinguish between a quiescent and an activated state among the CSCs population.

When applied to U87shTrkC cell fractions, inhibition of TrkC signaling appeared to favor a less proliferating F1 sub-population and a less clonogenic F3 sub-population, and the two fractions were similarly resistant to temozolomide-induced apoptosis (data not shown).

Cell culture and transfection

U87-MG and LN18 GB cell lines were purchased from American Type Culture Collection (ATCC). U87shBeclin1 and U87pLKO cell lines were provided by Dr Kögel and U87shATG5, U87shTrkC, LN18pLKO, LN18shATG5 cell lines were generated by transfection of U87-MG or LN18 cells with plasmids containing appropriate shRNAs (gift from Dr Priault, UMR CNRS 5095, Bordeaux). Four different clones were tested for ATG5 silencing (C163, C165, C166 and C167). Cell cultures were realized in Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, Life Technologies, Cergy-Pontoise, France), supplemented with 10% Fetal Bovine Serum (IDbio, Limoges, France) while transfected cells were grown in DMEM-FBS 10% containing 1 μg/mL puromycin (Fisher Scientific, Illkirch, France); both media contained GlutaMAX™. Cells were cultured under normoxia (5% CO₂, 20% O₂) or hypoxia (5% CO₂, 1% O₂), in a humidified incubator (Binder, Gennevilliers, France) for 96 h, seeded at a density of 10⁴ cells per cm² and 2.5.10⁶ cells/mL suspensions are used for SdFFF cell sorting or monitoring.

Treatment with the chemotherapeutic agent temozolomide was performed at 1.5 μM upon 48 h after the 96 h culture or after cell sorting by SdFFF. Chloroquine treatments (40 μM, accordingly to previous study, 32) were only performed during 24 h after 96 h of culture, 2.5.10⁶ cells/mL suspensions were used for SdFFF monitoring.

BrdU proliferation assay

Cell proliferation was tested with BrdU Cell Proliferation Assay Kit (Cell Signaling, Leiden, The Netherlands). Cells were seeded with 1,000 cells in 96 wells plates and cultured for 96 h or 144 h (96h + 48 h of TMZ treatment). Cells were then incubated with BrdU for 3 h at 37° C, then isolated, treated and incubated with the requested antibodies according to the manufacturer's instructions. Incubation with the horseradish peroxidase substrate 3,3',5,5'-tetramethylbenzidine (TMB), during 30 min was followed by addition of a Stop solution, and finally the reading of absorbance at 450 nm with a Multiskan FC Thermo Scientific microplate photometer (Thermofisher Scientific, Illkirch, France).

Ki67 staining

U87-pLKO, shBeclin1 and shATG5-165 cell lines were grown on glass coverslips (14 mm diameter, Menzel-Glaser, VWR, France) in a 24-well plate at the density of 10 000 cells/cm². After 72h of culture, cells were fixed in paraformaldehyde 4% for 15 minutes at RT and then washed with PBS. Saturation and permeabilization steps were performed using PBS-BSA3%-0.1% Triton X100 for 30 minutes at RT. The primary antibody anti-Ki67 (Santa Cruz, sc-7846) was diluted at the appropriate concentration in PBS-BSA3% and incubated overnight at 4° C with gentle agitation and in humid atmosphere. After several washes in PBS, the secondary antibody (Alexa Fluor 594, Invitrogen) was diluted in PBS-BSA3% and incubated for 2h at RT and protected from light. After PBS washes, DAPI (Sigma, D9564) staining was performed and mounting was realized using fluorescence mounting medium (Dako) before analysis under epifluorescence (Eclipse E800, Nikon, Champigny sur Marne, France).

Cell death detection

This test is based on the quantitative sandwich-enzyme-immunoassay-principle (sandwich ELISA) using antibodies against nucleosomes (Cell Death Detection ELISA, Roche, Mannheim, Germany). Apoptosis is characterized by the splitting of the cell and the cleavage of DNA between nucleosomes, which is the most accessible part of the DNA, leading to the formation of mono or oligonucleosomes. Detection of nucleosomes allows the identification of apoptosis cell death. Cells were seeded at 10,000 cells per well and assay was realized according to manufacturer’s instructions. Absorbance was measured at 405 nm, using a Multiskan FC microplate photometer (Thermofisher Scientific, Illkirch, France).

Colony formation unit assay

This test shows the ability of tumoral cells to form colonies, which is a property of stem cells. Cells (SdFFF-sorted or not) were seeded in wells of 6-wells plate (1,000 cells per well) in semi-solid medium constituted of 3 layers. The first layer contained 1 mL of mixed PBS containing 1% agar and culture medium (v/v : 1/1), the second layer contained 2 mL of PBS-agarose (0.7%) and culture medium (v/v : 1/1) added with 1 volume of cell suspension containing 1,000 cells per mL and the third layer consisted of 1.5 mL of culture medium. Cells were then incubated at 37° C and the medium was replaced every 3 days. After 21 days, the cells were fixed with 4% PFA for 15 min, stained using a 0.1% Crystal Violet solution for 20 min and washed several times with PBS to clearly distinguish every colony. Colony number and size were determined using ImageJ software (National Institute of Health, Bethesda, MD, USA). Experiments were performed in triplicate.

Quantitative RT-qPCR

RNA extraction was performed on cell pellets using TRIzol® (Invitrogen, Illkirch, France). Pellets were dissolved in 500 μL of TRIzol®, then 100 μL of chloroform (Sigma Aldrich, Saint Quentin Fallavier, France) were added. Samples were shaken for 15 seconds and incubated at room temperature for 15 min., then centrifuged at 16,000 g for 15 min at 4° C. Upper aqueous phase was gently transferred in new tubes and the remaining phase was frozen at -20° C and kept for further protein extraction. RNAs were reverse-transcribed with the High Capacity cDNA Reverse Transcription kit, according to the manufacturer's instructions (Applied Biosystem, Illkirch, France). PCR runs were performed using Taqman® Fast Universal PCR Master Mix (Applied Biosystem, Fisher Scientific, Illkirch, France); gene expression was detected with StepONePlus® Real-Time PCR System (Applied Biosystems, Illkirch, France) and normalized to GAPDH. Results were expressed as Relative Quantification values: RQ= (2^-Δ∆Ct).

Western blotting and semi-quantification of proteins

Protein extraction has been performed with two techniques: RIPA for pellets of total cultures, or TRIzol® for pellets from SdFFF sorting.

RIPA: Cells were lysed for 30 minutes at 4° C with RIPA buffer (50 mM Tris-HCl pH 7.4, 1%NP-40, 0.5% Na-deoxycholate, 0.1% SDS, 150 mM NaCl, 2 mM EDTA and 50 mM NaF, supplemented, prior to use, with 1 mM PMSF). Cells were disrupted by sonication at 60 kHz, then centrifuged at 16,000 g for 20 min at 4° C. Protein concentration was measured with the Bradford protein assay (Sigma Aldrich, Darmstadt, Germany) and Bovine Serum Albumin standard, and absorbance was measured at 595 nm with a Multiskan FC photometer (Fisher Scientific, Illkirch, France).

TRIzol®: 150 μL of absolute ethanol were added to the remaining phase kept after RNA extraction and the samples were gently shaken, incubated for 2-3 min at room temperature and centrifuged 5 min at 2,000 g at 4° C. Proteins were precipitated with 400 μL of isopropanol, then the samples were incubated for 10 min at room temperature and centrifuged for 10 min at 12,000 g at 4° C. Supernatant was eliminated and pellets were washed with 95% ethanol containing 0.3 M guanidine hydrochloride. Pellets were then diluted in 1 mL of the washing solution, incubated for 20 min at room temperature and centrifuged at 7,500 g for 5 min at 4° C. Supernatant was removed and pellets were resuspended in 500 μl of absolute ethanol; samples were shaken, incubated at room temperature for 20 min and centrifuged at 7,500 g for 5 min at 4° C. Pellets were air dried for 10 min max and resuspended in 1X Laemmli buffer (1% SDS, 5% glycerol, 30 mM Tris-HCl pH 6.8) containing 8 M Urea to facilitate protein dissolution and denaturation. DC Protein Assay (BioRad, Marnes-La-Coquette, France) was used for protein determination, according to manufacturer’s instructions. BSA was used as standard and absorbance was measured at 595 nm, using a Multiskan FC photometer (Fisher Scientific, Illkirch, France).

Polyacrylamide gels were cast (10-15% acrylamide concentration, according to protein size) and loaded with 30 to 60 μg SDS-denatured protein per well. After separation, proteins were transferred onto a PVDF membrane (Polyvinylidene difluoride, Biorad, Marnes-La-Coquette, France). After being blocked with TBS-BSA 5%, membranes were incubated overnight at 4° C with primary antibodies, then 1 h at RT with HRP-conjugated secondary antibodies. Membranes were washed with 1X TBS-0.1% Tween20 solution, then revealed using Immobilon Western chemiluminescent HRP substrate (Millipore SAS, Molsheim, France) with G-box (Ozyme, Fisher, Bioblock Scientific, Illkrich, France). Measurement of band intensities were done with ImageJ software.

Flow cytometry

After 24 h treatment with 40 μM chloroquine, pLKO cells were detached from dishes with Versene (Gibco, Thermo Fisher, Illkirch, France). Cells were rinsed with PBS and 1 μL of ViobilityTM (Macs Miltenyi Biotech, Bergisch Gladbach, Germany) was added to each experimental test. Cells were twice rinsed in PBS and fixed 10 min in 4% PFA. Cells were rinsed again with PBS and permeabilized 30 min at 4° C in Perm buffer III (BD Biosciences, Le Pont de Claix, France). Cells were incubated 30 min at 4° C with anti-Oct4-FITC-conjugated antibody (Biolegend, Ozyme) or anti-CD133-BV421-conjugated antibody (Biolegend, Ozyme) or anti-GFAP-APC-conjugated antibody (Miltenyi Biotec) or anti-DCX-PE- conjugated antibody (BD Biosciences). Marked cells were rinsed in PBS and analyzed with a Cytoflex flow cytometer (ThermoFisher) and the Flowlogic software TM (Macs Miltenyi Biotech).

SdFFF cell sorting

SdFFF device

The SdFFF separation device used in this study derived from previously described apparatuses [47] as schematized in Figure 3A. The apparatus was composed of two 880 x 47 x 2 mm polystyrene plates, separated by a Mylar® spacer in which the channel was carved. Channel dimensions were 788 × 12 × 0.175 mm with two 50 mm V-shaped ends. The measured total void volume (channel volume + connecting tubing + injection and detection device) was calculated after injection and retention time determination of a non-retained compound (0.10 g/L benzoic acid (Merck Millipore, Lyon, France) solution, UV detection at 254 nm). Sedimentation fields were calculated as previously described [48] and were expressed in gravity units (1 g = 980 cm/s²). Elution signals were recorded at 254 nm with a SPD-20AV UV/VIS Detector (Shimadzu, Champ sur Marne, France) and a NI9211 (10 mV input) acquisition device (National Instruments France, Nanterre, France) operated at 3 Hz and connected to a PC computer by the means of a laboratory-developed Visual Basic software (VB pro, Ver 6.0, Microsoft Corp.).

SdFFF elution conditions for cell sorting and monitoring

The optimal elution conditions were determined experimentally and were: flow injection through the accumulation wall of 100 μL U87-MG, U87pLKO ± CQ 40 μM, U87shBeclin1 and U87shTrkC cell suspensions (2.5×10⁶ cells/mL); flow rate: 1.00 ± 0.02 mL/min; mobile phase: sterile PBS 1X (Gibco, Life Technologies, Cergy-Pontoise, France), pH = 7.4; external multi-gravitational field strength: 20.00 ± 0.02 g (365.5 ± 0.2 rpm). These conditions, optimal for cell separation [49] under the biocompatible hyperlayer elution mode [50], were essentially evidenced by i) the R_obs dependence (R_obs = void time /retention time : t₀ / t_r) to the flow rate and the external field strength, and ii) a value of the average cell elevation in the channel thickness "s", larger than the cell radius. By the use of the mean R_obs value, it was possible to calculate the average cell elevation "s" with the following equation: $R_{obs}=\frac{6s}{\omega }$ [51] where ω corresponds to the channel thickness (175 μm). These conditions were also implemented for monitoring the Time-dependent fractions, F, collected as described in Figure 3C to obtain the most pure fractions containing either more or less differentiated cells or cancer stem cells. The first fractions actually contain the largest and less dense cells, eluted after the the void peak and, at the end of the cell peak, the last fractions contain smaller and denser cells. To obtain a sufficient number of cells, consecutive SdFFF fractions were collected and gathered. Total Peak, TP corresponds to whole cell population eluted by SdFFF and is used as control for biological assays performed on sorted cells.

Cell size analysis

A 256 channel Multisizer II Coulter Counter (Beckman Coulter, Fullerton, CA) was used to determine the mean diameter of the cell population. Cells were diluted in Isoton® (Beckman Coulter) to a final volume of 15 mL. The counting conditions were: 500 μL sample volume, cumulating three successive assays. Results were displayed as the mean ± standard deviation for three independent experiments.

Statistical analysis

All experiments were performed at least 3 times or more. For small number of samples (less to 10) Kruskall Wallis analysis were done and for more than 10 samples ANOVA tests were performed. All analyses were realized using the PAST 2.17c software.