Human adult mesenchymal stem cells (MSC) are of great interest in regenerative medicine because they are able to differentiate into osteoblasts, adipocytes, chondrocytes, myocytes and even into neural cells (Pittenger et al, 1999; Muraglia et al, 2000; Woodbury et al, 2000; Bianco et al, 2001). They also maintain homeostasis and modulate immune reactions (Rasmusson, 2006; ; Ringden et al, 2006). Although MSCs are easily obtained from bone marrow (BM) by a simple puncture, it is commonly admitted that in vitro expansion is necessary prior to therapeutic use because of their scarcity. An assessment of the initial mesenchymal progenitor content would appear to be necessary in evaluating BM grafts, so as to establish the threshold for subsequent study of the relationship between input abundance and in vitro cell expansion, or for comparing different cell sources or the influence of donor characteristics, such as age. To date, there is no international consensus regarding the evaluation of native, un-manipulated, mesenchymal progenitor cells in BM. So far, only one test, the colony forming unit-fibroblastic (CFU-F) assay, has been admitted as the principal test for evaluating the mesenchymal cell content of fresh BM (Friedenstein et al, 1974). However, this assay is dependent on several factors: the media (Sotiropoulou et al, 2006), the selected fetal calf serum (FCS) used (Caterson et al, 2002), the input cell density (Colter et al, 2001; Sekiya et al, 2002) and the colony counting method. These variables make it difficult to compare different studies. To rapidly assess the quality of fresh BM, it is necessary to use a non-retrospective method, such as flow cytometry. Numerous surface antigens are expressed by expanded cells (Pittenger et al, 1999; Deans & Moseley, 2000; Minguell et al, 2001) but the expression profile of the fresh cells remains largely unknown (Deschaseaux et al, 2003; Gronthos et al, 2003; Boiret et al, 2005) We have previously described, in a limited series of fresh BM from femoral heads, CD45− CD14− cell subsets expressing CD73 or CD49a and containing all, or most, CFU-F (Boiret et al, 2005). In that study, the immuno-phenotyping procedure was optimized to identify the mesenchymal compartment and the results were compared with those of an adapted standard functional test (CFU-F assay). The aim of the present study was to propose a simple, reproducible method that combined flow cytometry with functional assays so as to quantify accurately the mesenchymal content of BM samples.
The scarcity of mesenchymal stem cells (MSC) in bone marrow (BM) has justified their ex vivo expansion before therapeutic use, but a method to evaluate the quality of initial mesenchymal content and track the modifications induced by graft processing has not yet been proposed. The aim of this study was to establish such a procedure. Flow cytometric and functional assay methods were modified to count CD45− CD14−/CD73+ subsets containing all MSC and used them to study BM from spongy bone (SB) and iliac crest aspirate (ICA). These methods detected the target subsets in all BM suspensions derived from SB (n = 154) and ICA, (n = 44) with a satisfactory correlation between immuno-phenotyping and functional tests by low-density plating. We noted a higher overall MSC frequency in SB cell suspensions but a lower plating efficiency of CD45− CD14−/CD73+ SB cells under standard culture conditions.
We propose a cell quality control on un-manipulated BM cell suspensions to quantify the mesenchymal compartment with regard to varying donor factors, such as age and sampling site, that influence expansion and define a therapeutic threshold value. Furthermore, we were able to confirm differences in plating efficiency and proliferative capacity between two BM origins.
Materials and methods
Human bone marrow cell source and isolation
All samples were collected with the patients’ informed consent, according to a protocol approved by the Regional Ethics Committee. Analytic procedures were performed on BM cells from spongy bone (SB) collected from the femoral head during hip arthroplasty (n = 160); SB was put into Iscove's modified Dulbecco's medium (IMDM; Cambrex Bio Science, Paris, France) containing 10% fetal calf serum (FCS; Hyclone, South Logan, UT, USA) and 1% EDTA (Cambrex Bio Science). It was fragmented by grips and chisels and adherent cells were collected after a 10-min incubation with collagenase (0·0125% Sigma-Aldrich Chimie, Saint-Quentin Fallavier, France) at 37°C. All suspensions were then filtered through a 100 μm nylon mesh (Falcon, Becton Dickinson, Le Pont-de-Claix, France) then washed in phosphate-buffered saline (PBS)/4% FCS. All analytical procedures were validated on normal BM cells obtained by ICA from autologous or allogeneic donors for orthopaedic (n = 33) or haematopoietic (n = 11) applications, respectively; leucocytes were counted and the cell suspensions were analysed directly by flow cytometry and CFU-F assay. In some cases, we assessed the effect, on the mesenchymal cell content, of purifying mononuclear bone marrow cells by Ficoll sedimentation (Histopaque 1077; Sigma Chemical, St Louis, MO, USA) or of erythrolysis with ammonium chloride (Stem Cell Technologies, Vancouver, BC, Canada) for 10 min on ice.
Flow cytometric analysis
We adapted our previously described flow cytometric analysis for the identification of mesenchymal progenitor-enriched subsets from SB by counting CD45− CD14−/CD73+ or CD45− CD14−/CD49a+ cells (Boiret et al, 2005). Un-manipulated BM cells were incubated with PBS/5% AB human serum for 15 min to prevent non-specific binding, then labelled with anti-CD45-fluorescein isothiocyanate (FITC) and anti-CD14-FITC monoclonal antibodies (mAbs) (RMO52 clone; Immunotech, Beckman Coulter, Roissy Charles de Gaulle, France) in combination with anti-CD73-phycoerythrin (PE) mAb (AD2 clone; Pharmingen, Beckman Coulter) or anti-CD49a-PE mAb (SR84 clone; BD Biosciences Pharmingen, Le-Pont-de-Claix, France) for 15 min at room temperature. In most experiments, an anti-CD34 phycoerythrin cyanin 5·1 (PC5)-conjugated (581 clone; Immunotech, Beckman Coulter) mAb was added. Cells were washed twice in PBS/4% FCS and re-suspended in PBS/4% FCS with 1 μg/ml propidium iodide (PI). Viable cells (PI−) were gated to create a morphological dot plot with forward scatter (FS) and side scatter (SS). Acquisition and analysis of events were performed on Epics Elite cytometer (Beckman Coulter).
Owing to the low frequency of mesenchymal cells, the PE immunoglobulin (Ig) isotype control adjustment was critical. Windows need to be set, based on staining with a combination of irrelevant mAb, so that any non-specific event should not be included in CD45− CD14−/CD73+ or/CD49a+ windows (Fig 1B). The fluorescence intensity of the AD2 and SR84 clones was high and easily distinguishable from non-specific fluorescence (Fig 1C and D). At last, by taking into account technique sensitivity (10−4) and the expected frequency of CD45− CD14−/CD73+ or CD45− CD14−/CD49a+ cells (about 0·01%), we noted that a minimal acquisition of 5 × 104 viable mononuclear cells (MNC) was necessary to count these populations. This procedure always detected the subset of interest.
With the aim of proposing a simplified standard procedure, we compared the anti-CD73 and anti-CD49a staining patterns. Together they comprised a similar cell subset, as we observed a strong correlation (r = 0·83; P < 0·01) between the percentages of the two cell subsets. However, in some cases we found a few CFU-F in the negative CD49a subset but not in the CD73 negative population (Boiret et al, 2005). Furthermore, even though the mean fluorescence intensity (MFI) on fresh bone marrow cells was roughly similar and clearly distinct from the isotypic control, we noted, after ex vivo cell expansion, a fourfold reduction of MFI of CD49a staining and the appearance of negative cells, which made it difficult to differentiate the positive subset and to quantify all expanded mesenchymal cells (Fig 1H). In aiming to use immuno-phenotyping for BM quality control, it appeared to us that we should use this same procedure before and after in vitro expansion, even though the size and structure of the input and output cells were very different, as demonstrated by a comparative use of calibrated flow check particles. We therefore preferred to use anti-CD73 mAb for fresh and expanded BM mesenchymal cells. Preliminary experiments showed that most CD45− CD14− CD49a+ cells and all CD45− CD14− CD73+ clonogenic mesenchymal progenitor cells were present in mononuclear cell areas (data not shown). With the aim of locating progenitor cells in a FS/SS dot plot, CD45− CD14−/D73+ cells (n = 8) were sorted by gating monocyte-like cells and lymphocyte-like cells and subjecting these cells to the CFU-F assay.
In parallel, the CD34+ cell content was assessed to evaluate the quality of bone marrow. This was done using the routine International Society of Hematotherapy and Graft Engineering (ISHAGE) protocol with antiCD34-PE mAb (581 clone; Immunotech, Beckman Coulter), combined with antiCD45-FITC mAb (J33 clone; Immunotech, Beckman Coulter) and 7-Amino-Actinomycin D (7AAD), as well as a viability marker on Epics XLTM (Beckman Coulter). To evaluate BM dilution by peripheral blood, CD3+ cell frequency was determined in BM and in blood after staining with antiCD3-PE mAb (clone UCHT1; Immunotech, Beckman Coulter) associated with antiCD45-FITC mAb (J33 clone) and 7AAD, and after evaluating on Epics XLTM (Beckman Coulter).
Cells were seeded in 25 cm2 culture flasks (Falcon, Becton-Dickinson) at 2 × 104 to 40 × 104 cells/cm2 in basic mesenchymal medium, consisting of IMDM (Sigma®, Saint-Quentin Fallavier, France), 10% FCS (Hyclone), l-glutamine (2 mmol/l) and 1% Ciprofloxacin at 37°C under 5% CO2 in a water-saturated atmosphere. FCS was previously tested and approved by the Graft Group of the French Society of Bone Marrow Transplantation and Cell Therapy (Société Française de Greffe de Moelle-Thérapie Cellulaire, SFGM-TC). CFU-F frequency was determined directly by microscopic examination of flasks at day 10 after May–Grunwald–Giemsa (MGG) staining. CFU-F were classified according to colony size (<25, 25–50 and >50 cells). Numbers of progenitor cells were normalized for 1 × 106 input cells. In a limited series (n = 7), we adjusted the number of seeded SB cells by plating one CD45− CD14−/CD73+ cell per cm2 to compare SB and ICA cells under strictly identical conditions.
The results are expressed as mean ± SEM. When necessary, numeric values were compared by bilateral paired or unpaired Student's t-test. A P-value of 0·05 or less was considered to be statistically significant. Correlations between two parameters were analysed by calculating Pearson's correlation coefficient.
The majority of CD45− CD14−/CD73+ CFU-F were contained in monocyte-like BM cells
Because CD73 identified all CFU-F in the CD45− CD14− cell subset (Boiret et al, 2005 and see Materials and methods), we used the anti-CD45, -CD14 and -CD73 mAb combination to further locate the clonogenic progenitors by sorting CD45− CD14− CD73+ cells from gated lymphocyte- and monocyte-like cells on a FS/SS dot plot (n = 8) (Fig 2A). After MGG staining, the sorted cells showed morphology similar to that described elsewhere (Prockop et al, 2001). The monocyte-like cells showed only subtle differences (slightly increased size, a little more plentiful cytoplasm) when compared with lymphocyte-like cells (Fig 2). After 10 d in culture, most (92 ± 3%) CD45− CD14−/CD73+ clonogenic cells were monocyte-like cells (Fig 1B); we confirmed previously reported clonogenicity of around 2% (Boiret et al, 2005), corresponding to a 2- to 3-log enrichment when compared with the whole collected BM cell suspension. However, there was no difference in the initial proliferative capacity, as assessed from the distribution of colony size, between CFU-F from monocyte-like cell gating and those from lymphocyte-like cell gating (Fig 2C). With the aim of identifying a subset containing all CFU-F, we decided to gate all mononuclear cells. The gating sequence was conditioned by propidium iodide (PI) negative viable cells, followed by mononuclear cell gating and, finally, the number of CD45− CD14−/CD73+ subsets was counted.
CFU-F assay at low density correlated with immuno-phenotyping
Fresh BM cells were seeded in a simple, cytokine-free medium in the presence of FCS that had been selected for CFU-F proliferation. In these conditions, cell density is an essential parameter (Sekiya et al, 2002) but no consensus exists in the literature concerning the un-manipulated cell density required. For un-manipulated cells, we observed the previously reported inverse correlation between the input cell density and the progenitor cell frequency from Ficoll-separated cells (Digirolamo et al, 1999; Colter et al, 2000) (Fig 3A). There were more large colonies (>50 cells) in high seeding density cultures (Fig 3B) and a decrease in CFU-F between D7 and D10 (data not shown), suggesting that, in these conditions, colonies must fuse, with progenitors being underestimated. In these simple culture conditions, the optimal density was 2 × 104 un-manipulated cells/cm2. The results of the CFU-F assay showed that only a density of 2 × 104 cells/cm2 established a correlation (r = 0·56; P < 0·0001) with the phenotype, which is another reason for choosing this cell density (Fig 3C and D). However, in the context of cell quality control, we have to ensure that a functional test, which allows an a posteriori evaluation, can be adapted to the initial mesenchymal content. Therefore, given the heterogeneity of samples, we propose that the principles defined for quality control of haematopoietic cells should be applied, i.e. using two initial densities for seeding a sample of 2 × 104 cells/cm2 and one of 6 × 104 cells/cm2, to give a density that is three times higher so as to be certain of detecting progenitors in even the poorest samples. We used these assays to check that volume reduction or steps, such as erythrocyte lysis or sedimentation on Ficoll during BM processing, did not cause any selective loss of mesenchymal cells (data not shown).
BM mesenchymal cell content differed between spongy bone and iliac crest
After adjusting the analytical procedures using SB from femoral head, we applied them to ICA collected for cell therapy. At the same time, we hoped to confirm the reported influence of BM origin on mesenchymal cell content. The haematopoietic content evaluated by CD34+ frequency was 2·6-fold higher in SB than in ICA (P < 0·001) (Table I). It should be noted that the CD34+ cell count, determined by the addition of PC5-conjugated mAb to CD45CD14/CD73 staining, correlated strongly (r = 0·91; P < 0·0001) with results obtained using the reference method (ISHAGE) (Sutherland et al, 1996). In parallel, the comparison of the percentage of circulating CD3+ cells and bone marrow CD3+ cell frequency, which theoretically indicates the dilution of bone marrow by peripheral blood (16 ICA suspensions and 55 SB samples), showed a higher percentage of T cells in 33% of SB samples which were practically not diluted by peripheral blood. We found no significant inverse correlation between CD34+ and CD3+, and in our experience this parameter is a poor indicator of peripheral blood dilution and cannot explain the apparent heterogeneity of BM mesenchymal content.
|SB (n)||ICA (n)|
|MNC%||36 ± 0·9 (157)||26 ± 1·1 (44)|
|CD34+%||2·5 ± 0·11 (157)||0·98 ± 0·07 (44)|
|CD45− CD14−/73+%||0·15 ± 0·02 (154)||0·006 ± 0·012 (44)|
|CFU-F No/1 × 106 input cells 2 × 104 cells/cm2 seeding||295 ± 23 (51)||26 ± 10 (18)|
We detected CD45− CD14−/D73+ cells and CFU-F in all cases. The frequency of CD45− CD14−/CD73+ cells was about 25 times higher in SB than in ICA BM suspensions (0·15 ± 0·02% vs. 0·006 ± 0·0022%, P < 0·0001) (Table I). However, we did not observe any significant correlation between the percentages of CD34+ cells and CD45− CD14−/D73+, which refutes any strong relationship between the two BM compartments. Data from the CFU-F assay did not correspond exactly with the flow cytometric results. By taking account only of CFU-F from the optimal seeding density (2 × 104 cells/cm2), the frequency of CFU-F per 1 × 106 input cells was also higher in SB cells (295 ± 23 (4–736) vs. 26 ± 10 (6–110); P < 0·0001) but with only an 11-fold higher proportion of clonogenic progenitors, revealing a lower plating efficiency of CD45−/CD14−/CD73+ SB cells. We also noted fewer large colonies (>50 cells) when compared with CFU-F from the iliac crest (Table I). However, we calculated that between one and 30 CD45− CD14−/D73+ cells per cm2 from SB and ICA, respectively, were actually deposited in flasks. Therefore, the apparent lower plating efficiency could result from the higher density of progenitor input. In a limited series, we seeded all of the nucleated BM cells in order to deposit one CD45− CD14−/D73+ cell per cm2, whatever the BM source. In these conditions, we still observed a lower plating efficiency for SB cells (n = 6, data not shown).
This study has proposed a simple procedure that combines immuno-phenotyping and the CFU-F assay to quantify the mesenchymal content of BM samples. Although the expanded cells were well characterized (Deans & Moseley, 2000; Vogel et al, 2003), the phenotype of native MSC remains poorly known (Deschaseaux et al, 2003; Gronthos et al, 2003). In a small BM series, we identified two subsets – CD45− CD14−/CD73+ and CD45− CD14−/CD49a+– that contain all or most mesenchymal progenitors (Boiret et al, 2005). We showed that most clonogenic mesenchymal progenitors were located in the gated area of monocyte-like cells and some remaining progenitor cells in lymphocyte-like cells and confirmed the cloning efficiencies and CFU-F enrichment previously reported (Boiret et al, 2005). To demarcate a cell subset containing all CFU-F, we gated for all MNC. With the aim of applying the same immuno-phenotype before and after expansion, we propose the use of mAb against CD73, whose expression is mostly higher than 10 MFI value and clearly distinct from the Ig isotype control, because CD49a is downregulated during ex vivo expansion and, in our tests, was not expressed on all CFU-F (Boiret et al, 2005). We did not test STRO-1 mAb (Simmons & Torok-Storb, 1991), another potentially interesting antibody, because of the difficulty of including this stain with the other mAb used in our combination and to clearly identify the positive cell subset (Gronthos & Simmons, 1995; Gronthos et al, 2003). Moreover, this antigen is lost early during culture and prevents the application of the same immuno-phenotype before and after ex vivo expansion. In parallel, we performed functional analyses of the same cell population, i.e. on the raw BM cell suspension whereas, in the literature, CFU-F assay is generally carried out on Ficoll-separated cells. Furthermore, heterogeneous techniques have been described for cultivating and counting CFU-F, using different basal media, with or without growth factors and different incubation times, with culture reading performed from D7 to D15, or by identifying macroscopic colonies (Digirolamo et al, 1999; Colter et al, 2001). We propose here a simple, accessible and economic technique. As described elsewhere for expanded mesenchymal cells (Digirolamo et al, 1999; Colter et al, 2000; Sekiya et al, 2002), an inverse correlation was observed between seeded, un-manipulated, fresh BM cell density and the frequency of CFU-F normalized by an arbitrary number. 2 × 104 fresh BM nucleated cells/cm2 were used as the optimal cell density as a significant correlation was observed between immuno-phenotyping and CFU-F frequency even though higher cell density seemed to favour colony fusion and produced no significant correlation between pheno-typically and functionally defined subsets, an argument strengthening the choice of input cell density.
The frequency of CD34+ cells, which is usually considered to be an indicator of the quality of BM harvest, was not significantly correlated to the frequency of CD45− CD14−/CD73+ cells. This absence of any relationship between immature haematopoietic and mesenchymal cells suggests that the abundance of the two BM compartments was not directly reliable and that the variability in the purity of BM collected may not totally explain the BM heterogeneity for mesenchymal content. In theory, haemo-dilution could have been assessed from the proportion of CD3+ cells, but this parameter could not be included because there was a higher percentage of CD3+ cells in the SB from femoral heads, only slightly haemo-diluted, than in blood, possibly explained by the presence of lymphoid nodules in SB collected from older patients (Silver, 1969; Mazo et al, 2005) or by the inflammatory context of osteoarthritis impairment. We did not observe any inverse correlation between the frequency of CD3+ cells and that of CD34+ cells. Therefore only CD34+ cell count was finally included as the parameter for the evaluation of BM quality, which can be carried out in combination with CD45/CD14/CD73 labelling.
Bone marrow heterogeneity is well documented, including from different sampling sites conditioning the mesenchymal progenitor cell seeding density and consequently, the range of in vitro cell expansion (Digirolamo et al, 1999; Suzuki et al, 2001; Ahrens et al, 2004; Risbud et al, 2006). Here, analysis of BM suspensions showed a higher frequency of CD45− CD14−/CD73+ in SB of femoral heads than in BM from ICA. These results are consistent with the reported higher frequency of progenitors in SB-adherent cells from different sites (Suzuki et al, 2001; Ahrens et al, 2004; Risbud et al, 2006) than BM aspirates from iliac crest, suggesting either that mesenchymal progenitors are distributed heterogeneously in bone tissue, or that progenitor harvest differs according to harvesting method (aspiration through a trochar versus detachment of cells adherent to spongy tissue) (Lee et al, 2003). However, the plating efficiency of the isolated SB cells, as seen from the CFU-F/CD45− CD14−/CD73+ cell ratio, appeared lower than that of the BM cells from the iliac crest, and the distribution of colony size showed fewer large colonies from SB, suggesting functional differences between the two BM sources. However, these observations could be related to donor age [median: 67 years (SB) vs. 49 years (ICA)], or to contamination of the cell suspension with more mature mesenchymal cells detached from SB or to an artefact induced by lower seeding density of mesenchymal progenitors from ICA. Standardization by seeding one CD45− CD14−/CD73+/cm2 culture showed similar plating efficiency as the 2 × 104 BM cells/cm2, demonstrating that this input density was optimal.
For most clinical protocols, MSCs are first expanded in vitro mainly because of their scarcity in BM (Lazarus et al, 1995; Horwitz et al, 1999; Koc et al, 2000; Horwitz et al, 2002; Le Blanc et al, 2004; Bang et al, 2005). In most studies, the parameter most often taken into account is the total number of expanded mesenchymal cells. However, it is now necessary to be able to assess fresh BM mesenchymal content for several reasons. First, the question of systematic ex vivo expansion of mesenchymal cells before use is debatable. Indeed, the limited mesenchymal cell life span, telomere shortening and decrease in multi-lineage differentiation potential during ex vivo expansion indicate that there are modifications in the biological properties of native MSC during expansion (Banfi et al, 2000, 2002; Baxter et al, 2004). At the moment, it is not possible to state that expanded cells compensate for the properties lost or modified during in vitro culture by their numbers, but the study of un-manipulated mesenchymal cells means that it is necessary to count them. In an allogeneic strategy only, a short culture period is interesting to purify the MSC by eliminating haematopoietic cells involved in the graft-versus-host disease. Secondly, the evaluation of BM mesenchymal cell content is necessary to explain the extent of ex vivo expansion and the final cell harvest, as the initial MSC content of the BM is heterogeneous and inversely correlated to age (Muschler et al, 2001). Thirdly, the effect of each medium component on biological MC properties (Berger et al, 2006; Dimarakis & Levicar, 2006; Sotiropoulou et al, 2006) can be assessed only by comparing expanded cells with native cells, which implies that we must first be able to identify and quantify clonogenic un-manipulated progenitor cells. Despite many studies on these cells, graft quality assessment remains based on the CFU-F assay (Friedenstein et al, 1974), an approximate, non-standardized technique that is poorly reproducible (Owen, 1998; Phinney et al, 1999; Bianco et al, 2001) with response times that enable a posteriori evaluation only. The best way to assess mesenchymal content should associate rapid and standardizable quantitative and functional tests.
To conclude, we propose the identification of CD45− CD14−/CD73+ cells, produced by CFU-F assay at 2 × 104 and 6 × 104 cells/cm2, as standard procedures for evaluation of the mesenchymal content in fresh BM. Finally, the study of native mesenchymal cells is essential to reach several objectives: (i) Evaluation of graft quality at collection of BM cell suspension. (ii) Defining predictive parameters of therapeutic efficacy for grafts used without expansion. In some applications, such as orthopaedics, it may be possible to use a limited amount of native mesenchymal progenitor cells. Only Hernigou et al (2005) have proposed a progenitor threshold evaluated on fresh nucleated BM cells, for bone reconstruction using a functional test that remains imprecise and difficult to standardize (Hernigou et al, 2005). Furthermore, this group used an initial density of 8 × 104 cells/cm2, which, in our tests, was not an optimal seeding cell density. In this domain, immuno-phenotyping presents obvious advantages in terms of accuracy and reproducibility, and should make it possible to evaluate the amount of BM required for strategies requiring only a small amount of bone marrow. (iii) Evaluation of the effects of ex vivo manipulation of BM on the cell population of interest. All the steps of in vitro manipulation were likely to modify mesenchymal cell biology. Even though expanded MSC have been used in clinical applications of cell therapy, study of the effects of in vitro expansion on MSC properties is impaired by the lack of knowledge concerning native MSC. Therefore, an accurate evaluation of mesenchymal content in fresh BM is necessary.
This study was supported in part by grants from the Anti-Cancer League Regional Committee (Puy-de-Dôme) (Comité Départemental (Puy-de-Dôme) de la Ligue Contre le Cancer), Hospital-based Clinical Research Program (# UF 8153) and the Scientific Board of the French Blood Institute (Conseil Scientifique de l'Etablissement Français du Sang) (# 2005·11). The authors would like to thank Dominique Chadeyron for preparing the manuscript and Dr B. Aublet-Cuvelier for statistical analysis assistance.