Human Fetal Liver: An In Vitro Model of Erythropoiesis

We previously described the large-scale production of RBCs from hematopoietic stem cells (HSCs) of diverse sources. Our present efforts are focused to produce RBCs thanks to an unlimited source of stem cells. Human embryonic stem (ES) cells or induced pluripotent stem cell (iPS) are the natural candidates. Even if the proof of RBCs production from these sources has been done, their amplification ability is to date not sufficient for a transfusion application. In this work, our protocol of RBC production was applied to HSC isolated from fetal liver (FL) as an intermediate source between embryonic and adult stem cells. We studied the erythroid potential of FL-derived CD34+ cells. In this in vitro model, maturation that is enucleation reaches a lower level compared to adult sources as observed for embryonic or iP, but, interestingly, they (i) displayed a dramatic in vitro expansion (100-fold more when compared to CB CD34+) and (ii) 100% cloning efficiency in hematopoietic progenitor assays after 3 days of erythroid induction, as compared to 10–15% cloning efficiency for adult CD34+ cells. This work supports the idea that FL remains a model of study and is not a candidate for ex vivo RBCS production for blood transfusion as a direct source of stem cells but could be helpful to understand and enhance proliferation abilities for primitive cells such as ES cells or iPS.


Introduction
The cell therapy approach which consists of generating cultured RBC (cRBC) in vitro after amplification of stem cells (SC) makes sense in a context of chronic difficulties of obtaining blood supplies. More than blood supplies, hematopoietic differentiation of human stem cells into RBC has important therapeutic implications, including the possibility to produce virus-free units or tailor-designed phenotypes for transfusion purposes. Our team has established an experimental procedure to reproduce in vitro terminal erythropoiesis from adult hematopoietic stem cells (HSC) from diverse sources (peripheral blood (PB), bone marrow (BM), and cord blood (CB)) [1,2]. This protocol, using appropriate cytokines and a specific microenvironment (including stromal murine MS5 or Mesenchymal Stem Cells (MSC)) in a serum-free medium, allows HSC proliferation and terminal differentiation in mature and functional enucleated RBC containing adult haemoglobin [2]. Major advances have allowed towards the in vitro production of RBCs from diverse sources in a few years. Nevertheless, their transfusional future will become a reality only if we are able to produce functional transfusable RBCs at a large scale. This will of course require not only the conception of appropriate industrial tools, but also finding the best source of stem cells. To date, the most accessible and proliferative source of HSC in a quantitative aspect is cord blood (CB) [3].
However, this cell source is dependent on donations and ex vivo production of RBC is restricted to a system of production in batches, using the finite quantity of HSC available in a CB unit. Such are its limits.

Stem Cells International
Consequently, our present efforts are focused on establishing conditions to produce RBC using a permanent and inexhaustible source of stem cells. Human embryonic stem cells or recently discovered induced pluripotent stem cells (iPS) [4,5] are the natural candidates. As other teams, we have shown the possibility to reconstitute in vitro erythropoiesis starting from very primitive stem cells [6][7][8][9][10][11][12][13], and we were the first to report the erythroid differentiation and maturation of iPS lines (from fetal and adult fibroblasts) into mature enucleated RBC (4% to 10% compared to 52% to 66% from human ES cell line (H1)), which synthesized functional fetal hemoglobin [9].
The demonstration that primitive cells can be differentiated in vitro to cells similar to their natural counterpart was the first step for a potential therapeutic application. But there is still a major problem for a hypothetical application in transfusion: their amplification. Indeed, in our conditions, 1 ES or 1 iPS cell can generate 5000 or 1000 RBC, respectively, [9] whereas 1 CD34 + from CB can give up to 2.5 × 10 5 RBCs [2]. This crucial difficulty of amplification of ES or iPS has not yet been solved by any team.
Rollini et al. have demonstrated that fetal liver (FL) could be an alternative source of HSC despite a limited total cell number per tissue, due to their high proliferative capacity [14]. Indeed, the site of erythropoiesis migrates during development. It takes place first in the yolk sac and the para-aortic region. Erythropoiesis then migrates to the fetal liver between the 8th and the 22nd week of gestation to finally take place in the BM [15]. In the present work we investigated the capacity of CD34 + cells isolated from human FL, an ontogenic source intermediate between very primitive cells (ES and iPS) and adult stem cells, to expand and differentiate into erythroid lineage cells.
We cocultured FL-derived CD34 + cells over an in vitro reconstituted bone marrow (BM) microenvironment. As observed for hES or iPS, only a fraction of erythroid cells reached terminal differentiation into mature RBC (15%). These few RBCs contained nearly exclusively fetal hemoglobin at the protein level. However, as compared to adult HSC, we could observe a dramatic expansion of FLderived erythroid cells, 100-fold higher when compared to CB CD34 + .
FL is an alternative model that might be relevant to understanding mechanisms both implied in amplification and differentiation of HSC in erythropoiesis and could be used as a model for other differentiation pathways.

Fetal Liver.
Fetal livers were surgically obtained from 6 aborted fetuses (after 12-14 weeks of amenorrhoea) after informed consent of patients was given. All experiments and procedures were in agreement with the guidelines from the French Health ministry and the Agence de la Biomédecine. FL was dissociated within less than 4 hours after collection and flushed through a 70 μm filter and preserved in a PBS/BSA 10% (v/v) (Sigma Aldrich, Lyon, France) solution until CD34 cell immunomagnetic separation, which was performed within a maximal 1 hour delay.

Generation of Red Blood Cells Ex
Vivo. CD34 + cells were cultured in a three-step procedure in a serum-free medium as previously described [2] with some modifications such as the source of serum or the use of FL mesenchymal stem cells (MSCs) as stroma. 2 × 10 4 CD34 + cells/mL were first cultured in IMDM medium (Biological Industries, ATGC Biotechnologies) supplemented either with 1% BSA or 1% human AB plasma, 10 μg/mL insulin (Sigma), and 120 μg/mL iron-saturated human transferrin (Sigma). In the first step (days 0-8), 2 × 10 4 /mL CD34 + cells were cultured in the presence of 10 −6 M hydrocortisone (Sigma), 100 ng/mL stem cell factor (SCF, Peprotech, Neuilly sur Seine, France), 5 ng/mL IL-3 (R&D Systems), and 3 IU/mL erythropoietin (Eprex, kindly provided by Janssen-Cilag). On day 4, one volume of cell culture was diluted in four volumes of fresh medium containing hydrocortisone, SCF, IL-3, and erythropoietin. In the second step (days 8-10), the cells were replated at 5 × 10 4 /mL and cocultured on an adherent stromal layer in fresh medium supplemented with erythropoietin. In the third step (days 10-18), cells were cultured on an adherent stromal layer in fresh medium without cytokines. The cultures were maintained at 37 • C in 5% CO 2 . The adherent cell layer consisted of either the MS-5 stromal cell line (provided by K. Mori) or fetal mesenchymal stromal cells (MSCs) established from the CD34 negative fraction of FL liver samples.

Cells and Hemoglobin Characterisation
Hematological Staining. Cell differentiation was monitored throughout culture by morphological analysis of the cells after cytocentrifugation and May-Grünwald-Giemsa (MGG) staining.
Hemoglobin Analysis. The pattern of hemoglobin (Hb) synthesis was analyzed by high-performance liquid chromatography (HPLC) (Biorad variant II, BioRad Laboratories, Munich, Germany) with a high-resolution betathalassemia program.
Total RNA was prepared from cells collected at different times of RBCs culture (days 8,11,15,18). RNA was extracted using the Trizol method according to the manufacturer's instructions (Invitrogen, Paisley, Scotland). 1 μg of DNasetreated RNA was transcribed into cDNA using 200 units of SuperScript II reverse transcriptase (Invitrogen) and 150 ng of random primers (Invitrogen). The resulting cDNA was aliquoted to avoid repeat freeze/thaw cycles. Real-time PCR was performed using an ABI PRISM 7700 Sequence Detection System (Applied Biosystems, Foster City, Calif, USA). The 18S gene was used as an internal reference. All primers and probes were from Applied Biosystems.
Other genes were amplified using SYBRgreen chemistry, and the GAPDH gene was used as an internal reference. All primers were from Eurogentech (Liege, Belgium).
For both chemistries, PCR were performed in duplicate using either the TaqMan Master Mix or SYBR green Master Mix with 10 ng of cDNA and 300 nM of primers in a final

Massive Expansion and Commitment of Human Erythroid
The development of a large and almost pure erythroid cell population can probably be attributed to the preferential proliferation of the generated erythroid progenitors and their subsequent differentiation and maturation. On day 3, 100% of the cells were clonogenic, most belonging to the erythroid lineage (44% BFU-E, 7% CFU-E, 33% CFU-GM and 16% CFU-mix). On day 8, 5% of the cells were clonogenic, 90% of these being BFU-Es. Late erythroid CFU-E progenitors proliferated extensively and constituted 57% of the clonogenic cell population by day 10 (Figure 2). Cells in culture lost their immaturity as early as day 8 as assessed by CD34, CD117, and CD33 negativity, while their erythroid commitment was confirmed by CD235a, CD71, and CD36 positivity. Day 18 cultured RBCs were CD235a + . These cells still expressed CD36 and CD71 markers, as was also observed in our model using CB and BM sources (Figure 3(a)).
On an MS5 stromal layer, FL CD34 + cells displayed the same erythroid maturation kinetics as CD34 + cells from other sources (CB, BM, or peripheral blood, data not shown) up to day 15. However, their erythroid maturation was blocked by day 18 at the acidophilic stage, and very little nucleus expulsion was observed (3 to 15%) (Figure 3(b)). On the hypothesis that a fetal microenvironment would provide a more suitable niche to achieve complete maturation, we established cocultures over autologous or allogeneic MSCs grown from the FL CD34 negative fractions, but with no improvement in the enucleation efficiency even by extending culture up to day 24.

Hemoglobin Analysis of FL-Derived Erythroid Cells.
Quantitative real-time PCR analysis of total RNA at different times of culture (days 8, 11, 15, and 18) showed that, except for the β globin gene whose expression is constant regardless of the stage of the culture, the expression level of mRNA coding for the other globin genes is upregulated during the erythroid differentiation process. More particularly, the globin gene expression of the FL-derived erythroid cells showed to be mainly embryonic (ζ and ε globin chains) and fetal (γ globin chains). No adult β globin gene expression could be detected. BM-derived erythroid cells showed an adult globin profile as they mainly expressed β globin chains and to lesser extent fetal (α, γ) globin chains. No embryonic genes (ζ-and ε-globin chains) could be detected. CB-derived erythroid cells showed a similar gene profile but had a stronger expression of the fetal γ-globin chains; embryonic genes (ζ-and ε-globin chains) could be detected but with a variable and low expression (Figure 4(a)). HPLC analysis (n = 4) performed on erythroid cells derived from FL revealed mainly fetal hemoglobin (Hb) regardless of the time of culture. On day 8, the cells contained 68% (59-77%) fetal hemoglobin (HbF) including an acetylated fraction. A very early migrating subfraction was detected and could be identified as Bart Hb, which corresponds to the association of four γ globin chains. At the end of culture (days [19][20][21][22][23][24], 86% (80-89%) of the cells contained HbF and less than 2% (1.2-1.8%) adult HbA (Figure 4(b)).

Gene Expression of Factors Involved in Definitive Erythropoiesis.
In an attempt to explain the blockage at the acidophilic stage of the FL-derived erythroid cells, we analyzed the expression pattern of factors known to be involved in definitive erythropoiesis. We compared PCR results with those obtained for BM and CB samples taken at the same time of erythroid culture (days 8,11,15,18). The 2 −ΔΔCt method was applied to establish relative expression of genes ( Figure 5). As compared to BM or CB samples, RT-PCR revealed a strong downregulation of all genes for FL-derived cells regardless of the stage of the culture. BM and CB had the same expression pattern of factors except for STAT5a factor (for which CB showed the same profile as FL). We also observed that the levels of gene expression were higher for BM than for CB.

Discussion
A large variety of ontogenic sources of HSC have been explored for cellular therapies. Because of constantly increasing demand in blood supplies, one particular field of interest is transfusion. Thus, we and others have described experimental procedures which allow the massive expansion of CD34 + cells from various sources and the generation of mature RBCs [1,2,[19][20][21][22]. To date, the most proliferative and accessible source of stem cells for generating cultured RBCs (cRBCs) is cord blood. It presents, nevertheless, the drawbacks of being restricted to voluntary donations and is not unlimited; 1 CB unit gives one batch of cRBC production.
Because they are an unlimited source of stem cells, many studies are being carried out with ES cells and more recently with iPS. Critical advances have allowed progress towards the in vitro production of functional RBC from these sources in a few years. However their limited ability to expand in vitro is a limitation to large-scale production.
In human development, FL is a rich source of hematopoietic progenitors between the 8th and the 22nd week of gestation as it is the exclusive site of erythropoiesis. Although "conventional" sources of HSC for generating cRBC include both adult (mobilized peripheral blood or CB) and primitive stem cells (ES or iPS cells), evidence suggests that FL may represent a rich alternative source of "early" HSCs, displaying multipotency, possessing high proliferative and repopulating potential [23][24][25][26]. Indeed, it has been proven to be efficient in transplantation [27][28][29].
In the present work, we addressed the question of the possibility of generating in vitro mature cells from HSC of such an intermediate ontogenic source (FL cells) between ES cells and adult cells. In our experimental conditions, FL cell maturation is mainly blocked at the acidophilic stage. Indeed, only few cells (3-15%) achieved terminal differentiation. We made the hypothesis that the blockade of maturation observed in vitro was the consequence of an inappropriate microenvironment. On the assumption that a microenvironment from the same ontogenic stage would better support erythroid differentiation, we cocultured FLderived CD34 + cells on FL-derived MSC instead of the murine MS5 stromal cell line. However, no difference was observed in terms of terminal maturation, and the produced cells were mostly nucleated, whereas in the same culture conditions adult stem cells could generate fully enucleated cRBC. When we analyzed by RT-PCR globin transcripts of FL-derived erythroid cells, the transcriptomic profiles were mainly fetal (γ-globin chains) and embryonic (ζ-and εglobin chains). In the same way, BM-derived erythroid cells showed an adult globin profile as they mainly expressed β-globin chain and to lesser extent fetal chains (α-, γglobin chains). No embryonic genes (ζ-and ε-globin chains) could be detected. In comparison, CB-derived erythroid cells showed a similar gene profile but had a stronger expression of the fetal γ-globin chains and embryonic genes (ζ-and ε-globin chains) could be detected but with a variable and low expression. By analyzing the globin genes expression, we could observe that our system of culture mimics what happens during development and leads to globin gene expression patterns in accordance with the cells' ontogeny.
The precise molecular mechanisms that regulate the erythroid differentiation process remain to be fully elucidated despite many genes and pathways have been described to be essential in erythropoiesis. In order to understand the reason for the enucleation deficiency observed with FL cells, we analyzed by RT-PCR the expression of some genes known to be involved in definitive erythropoiesis (i.e., GATA-1, FOG 1, SOX6, GFI 1b, STAT5a, FOXO3a) upon the following rationale. GATA-1 is found exclusively in hematopoietic cells and is particularly important in erythroid and megakaryocytic lineages. In vitro, in the absence of GATA-1, a maturation deficit can be observed, indicating its crucial role in terminal erythroid differentiation [30]. FOG-1-null embryos present a defect both in primitive and definitive erythropoiesis with a blockage of maturation at the stage of proerythroblast [31].
The signal transducer and activator of transcription STAT5 is responsible for the control of proliferation, differentiation, and apoptosis, via its effect on gene expression. STAT5 proteins [32] are suggested to play an important role in hematopoiesis, but there are numerous contradictory reports on the role of STAT5 in normal hematopoietic cell development. A recent study [33] shows that overexpression of an activated mutant of (STAT5) induces erythropoiesis and increases long-term proliferation of human hematopoietic stem/progenitor cells. siRNA silencing of FOXO3a as well as the inactivation of GFI 1b led to erythroid differentiation blockage [34,35]. Finally, it has been reported that, in CD34 + from CB, SOX6 accelerated the kinetics of erythroid maturation and increased the number of cells that reached the final enucleation step [36].
Our results clearly show that, in our culture conditions, all the genes we studied are strongly downregulated in FL cells, whereas they are upregulated both in BM and CB cells, notably in BM cells. The pattern of gene expression that we observed was in concordance with the enucleation defect observed in FL but could not fully explain it. Indeed this molecular analysis is qualitative and not functional. Nevertheless, it suggests that genes involved in terminal differentiation are differently regulated in the function of the cell ontogeny and might at least in part explain the  differences observed in the ability to generate RBC in vitro. For many reasons, ES or iPS cells would be the natural candidates for the regenerative medicine of tomorrow. Through this model of erythropoiesis, one can understand how different cells behave according to their ontogeny. As it is titled in the publication of Williams et al. "Children are not little adults: just ask their hematopoietic stem cells" [37]. Extensive analysis on models explaining these differences would be of great interest to fully control the differentiation pathway of stem cells and would be helpful for the next generation of stem cell therapy.
Even if the generation of RBCs from ES cells or iPS has been quite routinely achieved by several teams [6][7][8][9][10][11][12][13], none succeeded in a 100% enucleated population whatever the culture conditions. Primitive cells seem to underperform in terms of production yield of fully mature cRBCs in vitro. The question whether this is due to their ontogeny or related to inappropriate culture conditions remains open.
If FL has no advantage in terms of final maturation over pluripotent stem cells, we observed an interest in terms of amplification. As far as the hematopoietic pathway is concerned, FL is known to contain a compartment of colony-forming cells with a high proliferative potential [14,[23][24][25]. These observations were confirmed by our findings of (i) a strong expansion of CD34 + FL cells and (ii) 100% cloning efficiency in hematopoietic progenitor assays after 3 days of erythroid induction, as compared to 10-15% cloning efficiency for adult CD34 + cells [17]. The proliferation capacity of stem cells from various sources is partly related to their ontogenic origin. ontogeny-related changes in proliferative potential of human hematopoietic cells in vitro have already been suggested in the literature, and analyses have shown, by comparing bone marrow, peripheral blood, umbilical cord blood, and fetal liver, that FL appears to represent a good target for ex vivo stem cell expansion [38,39].
We reproduced this observation in our culture conditions. Indeed, we observed a 100,000-and 25,000-fold expansion for cells from CB, BM, respectively. Strikingly, for FL-derived CD34 + cells, in the same culture conditions, the expansion reached more than 1.2 × 10 6 -fold.
FL remains a model of study and is not a candidate as a direct source of stem cells for ex vivo RBCs production for blood transfusion. On the contrary, because iPS cells can proliferate indefinitely, they are obviously the best candidate to set up complementary sources of RBCs for transfusion. But their clinical application will need a drastic increase of their ability to proliferate in vitro. The present observations relating to erythropoiesis from FL could pave the way for the design of new strategies to generate induced pluripotent stem cells. This idea is supported by the work of Kim et al. [40]. In a murine model, these authors observed that iPS harbor residual methylation signatures of their somatic tissue of origin, which favors their differentiation into lineages related to the donor cells while restricting alternative cell fates. According to this logic, we could imagine that a cell with a high proliferative capacity thereby would be optimal for the production of RBC.