Developmental changes in human megakaryopoiesis

Authors

  • O. Bluteau,

    1. Institut National de la Sante et de la Recherche Medicale, UMR 1009, Laboratory of Excellence GR-Ex, Villejuif, France
    2. Université Paris-Sud, Villejuif, France
    3. Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • T. Langlois,

    1. Institut National de la Sante et de la Recherche Medicale, UMR 1009, Laboratory of Excellence GR-Ex, Villejuif, France
    2. Université Paris-Sud, Villejuif, France
    3. Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • P. Rivera-Munoz,

    1. Université Paris-Sud, Villejuif, France
    2. Institut Gustave Roussy, Villejuif, France
    3. Institut National de la Santé et de la Recherche Médicale, UMR, Villejuif, France
    Search for more papers by this author
  • F. Favale,

    1. Institut National de la Sante et de la Recherche Medicale, UMR 1009, Laboratory of Excellence GR-Ex, Villejuif, France
    2. Université Paris-Sud, Villejuif, France
    3. Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • P. Rameau,

    1. IRCIV, Plateform of Flow Cytometry, Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • G. Meurice,

    1. IRCIV, Plateform of Functional Genomics, Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • P. Dessen,

    1. Institut National de la Santé et de la Recherche Médicale, UMR, Villejuif, France
    2. IRCIV, Plateform of Functional Genomics, Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • E. Solary,

    1. Institut National de la Sante et de la Recherche Medicale, UMR 1009, Laboratory of Excellence GR-Ex, Villejuif, France
    2. Université Paris-Sud, Villejuif, France
    3. Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • H. Raslova,

    1. Institut National de la Sante et de la Recherche Medicale, UMR 1009, Laboratory of Excellence GR-Ex, Villejuif, France
    2. Université Paris-Sud, Villejuif, France
    3. Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • T. Mercher,

    1. Université Paris-Sud, Villejuif, France
    2. Institut Gustave Roussy, Villejuif, France
    3. Institut National de la Santé et de la Recherche Médicale, UMR, Villejuif, France
    Search for more papers by this author
  • N. Debili,

    1. Institut National de la Sante et de la Recherche Medicale, UMR 1009, Laboratory of Excellence GR-Ex, Villejuif, France
    2. Université Paris-Sud, Villejuif, France
    3. Institut Gustave Roussy, Villejuif, France
    Search for more papers by this author
  • W. Vainchenker

    Corresponding author
    1. Institut National de la Sante et de la Recherche Medicale, UMR 1009, Laboratory of Excellence GR-Ex, Villejuif, France
    2. Université Paris-Sud, Villejuif, France
    3. Institut Gustave Roussy, Villejuif, France
    • Correspondence: William Vainchenker, INSERM UMR1009, Institut Gustave Roussy, 114 rue Edouard Vaillant, 94805, Villejuif cedex, France.

      Tel.: +33 1 42 11 42 33; fax: +33 1 42 11 52 40.

      E-mail: verpre@igr.fr

    Search for more papers by this author

  • See also Ghevaert C. Megakaryopoiesis through the ages: from the twinkle in the eye to the fully grown adult. This issue, pp 1727–9..
  • Manuscript handled by: S. Watson
  • Final decision: P. H. Reitsma, 10 June 2013

Summary

Background

The molecular bases of the cellular changes that occur during human megakaryocyte (MK) ontogeny remain unknown, and may be important for understanding the significance of MK differentiation from human embryonic stem cells (hESCs)

Methods

We optimized the differentiation of MKs from hESCs, and compared these with MKs obtained from primary human hematopoietic tissues at different stages of development.

Results

Transcriptome analyses revealed a close relationship between hESC-derived and fetal liver-derived MKs, and between neonate-derived and adult-derived MKs. Major changes in the expression profiles of cell cycle and transcription factors (TFs), including MYC and LIN28b, and MK-specific regulators indicated that MK maturation progresses during ontogeny towards an increase in MK ploidy and a platelet-forming function. Important genes, including CXCR4, were regulated by an on–off mechanism during development.

Discussion

Our analysis of the pattern of TF network and signaling pathways was consistent with a growing specialization of MKs towards hemostasis during ontogeny, and support the idea that MKs derived from hESCs reflect primitive hematopoiesis.

Introduction

The first and transient wave of hematopoiesis, so-called ‘primitive hematopoiesis’, begins in the yolk sac and leads to the production of erythroblasts, macrophages, and megakaryocytes (MKs) [1, 2]. A second wave of hematopoiesis (definitive hematopoiesis) emerges in the aorta-gonado-mesonephros region, and gives rise to all myeloid and lymphoid lineages [2]. In mammals, definitive hematopoiesis begins in the fetal liver, and migrates to the bone marrow at the end of gestation. This migration is associated with profound changes in the biology of human stem cells (HSCs), which proliferate when they colonize the fetal liver, and become quiescent in the bone marrow to prevent their exhaustion.

Ontogenic changes are observed in different lineages. They are sustained by variations in transcriptional networks, and may be accompanied by subtle changes in cell functions thoughout the course of ontogeny. Developmental changes in erythropoiesis are the best understood, and the successive switches in globin chain synthesis clearly distinguish primitive, fetal and adult hematopoiesis [2]. Functional changes related to ontogeny have also been observed in macrophages [3].

Changes in MK ploidy and platelet function do not recapitulate all of the ontogeny-associated changes in MK differentiation [4, 5]. Three observations suggest that embryonic and fetal MKs and platelets may have specific properties and functions. First, MKs/platelets may have functions in innate immunity and tissue remodeling through cytokine secretion and cell–cell interactions [6], as confirmed by the role of platelets in fetal lymphatic development [7]. Second, acute megakaryoblastic leukemia (AMKL) is a rare malignant disorder that is observed mainly in infants and young children, suggesting that the oncogenic events arise on a fetal progenitor [8, 9]. Finally, the thrombocytopenia–absent radii syndrome is an inherited platelet disorder that leads to profound thrombocytopenia in children and may resolve naturally in adults [10].

Here, we performed gene expression profiling of MKs from human fetuses, neonates, and adults. We also used MKs derived from human embryonic stem cells (hESCs), under the hypothesis that they are surrogates for human yolk sac-derived MKs [11-13]. These hESC-derived MKs (MKESs) were compared with MKs obtained by ex vivo differentiation of CD34+ progenitors collected from fetal liver samples (MLFLs), cord blood (MKCBs) and adults blood samples (MKADs), and finally with yolk sac-derived MKs (MKYSs).

Materials and methods

Cell lines and culture media

H1 (NIH code WA01) and H9 hESC (NIH code WA09) lines were obtained from the WiCell Research Institute, and HUES 1 and HUES 7 cell lines were obtained from D. A. Melton (Harvard Stem Cell Institute), and maintained in an undifferentiated state as previously described [12].

Blood samples

Informed consent was obtained in accordance with the Declaration of Helsinski. Blood samples from individuals after mobilization and human umbilical cord blood from normal full-term deliveries were collected after informed written consent had been obtained according to institutional guidelines (AP-HP and Agence de Biomédecine, St Denis, France).

Human embryos

A total of five human embryos were collected after maternal informed written consent had been obtained according to French guidelines (Agence de Biomédecine). Yolk sacs were dissected from two embryos aborted at 36–39 days, dissociated, and sorted for the collection of CD34/CD41+ primitive MKs (MKYSs). Fetal livers were dissected from three embryos aborted at 100–110 days, and cells were processed as described below.

MK differentiation from fetal liver, cord blood and adult CD34+ cells

CD34+ cells were isolated with an immunomagnetic bead technique (Miltenyi, Paris France), and grown in serum-free medium supplemented with human recombinant thrombopoietin (TPO) and stem cell factor (SCF) [14]. TPO and SCF were kindly provided by Kirin Pharma Company (Tokyo, Japan) and Biovitrum AB (Stockholm, Sweden).

Antibodies and flow cytometry

The conjugated mAbs used for flow cytometry experiments are listed in Table S1. Indirect immunofluorescence labeling was performed for studies of MPL expression (Amgen France, Neuilly sur Seine, France). Cells were analyzed and sorted as previously described [15], except for mature MKs, which were double-sorted on the CD41high/CD42high/CD14/CD15 phenotype.

MK differentiation of hESCs

hESCs were mechanically harvested and dissociated as small clumps, and seeded on OP9 cells in the presence of 20 ng mL−1 vascular endothelial growth factor (VEGF) (Miltenyi) and 20 ng mL−1 TPO [16]. At day 7, 25 ng mL−1 human SCF was added. At days 10–12, CD34+/CD41+ cells were sorted and were replated on OP9 cell monolayer in the presence of TPO and SCF.

Quantification of MKs bearing proplatelets

Sorted MKs from hESC lines at day 18 of culture or from fetal liver, cord blood and adult blood at day 10 of culture were plated in 96-well plates at a concentration of 2000 cells per well in serum-free medium in the presence of TPO (10 ng mL−1). Four days later, MKs displaying proplatelets were quantified by enumerating 500 cells per well underg an inverted microscope (Carl Zeiss, Le Pecq, France) at a magnification of × 200. MKs displaying proplatelets were defined as cells showing one or more cytoplasmic processes with constriction areas (three wells were examined for each condition).

Microarrays and bioinformatics analyses

Hybridizations were performed with a dye-swap strategy on whole-human-genome dual-color 8 × 60K oligonucleotide microarrays or on 8 × 60K human microRNA (miRNA) microarrays (Agilent Technologies, Massy France). Agilent feature extraction software (version 10.7.3.1) was used to quantify the intensity of fluorescent images and to apply a linear/Lowess normalization. All processing methods used for gene expression analyses were performed on the median signal from Agilent feature extraction raw data files, with functions and packages collected in the R Bioconductor project [17]. Normalization was performed with the ‘normalizeWithinArrays’ and the ‘normalizeInterArrays’ functions from r package limma followed by InterArrays [18]. The quality of the arrays was validated with the array quality metrics package from Bionconductor [19]. All intensity profiles were then imported into brb array tools software [20] for statistical analysis. The data were filtered at a minimal intensity of 10 and a minimal difference of 1.5-fold from the median in at least 20% of the arrays. Differentially expressed genes were identified among two or more classes with the class comparison method, a random-variance t-test (or F-test) with a minimal P-value of 0.001. Clustering experiments were performed with dchip software (version 1.0.0.1; Dana Farber Institute, Boston, MA, USA) on whole brb-filtered data or specific signatures [21].

For miRNA microarrays, control spots were systematically removed, and flagged spots (gIsFeatNonUnifOL and gIsSaturated columns from raw files) were considered as missing values. Quantile normalization was performed on the gMedianSignal values of each sample with the normalizeBetweenArray function from the limma package. To summarize the miRNA data, the mean of all probes for a given miRNA was computed, and the corresponding value was assigned to the miRNA. Data were filtered according to the maximum number of missing values allowed for each miRNA (15%): miRNAs with < 15% of missing values across all samples were kept, and the missing values were systematically replaced by use of the KNN algorithm from the impute.knn function of the impute r package. Otherwise, the miRNAs were discarded from analysis. To assess differentially expressed miRNAs, we first estimated the fold changes and standard errors between two groups of samples by fitting a linear model for each probes with the lmFit function. We applied empirical Bayes smoothing to the standard errors from the linear model previously computed with the eBayes function. We used the topTable function to extract a table of the top-ranked miRNAs or genes from the linear model.

The microarray data have been submitted to the Array Express data repository at the European Bioinformatic Institute under the accession numbers E-MTAB-1452 (whole genome microarray) and E-MEXP-3816 (miRNA microarray).

Global analysis of enrichment was performed with genecodis (specific stage signature) or gseav2.0 (whole brb-filtered data) to determine gene overrepresentation at each stage with gene ontology (GO) ‘Biological process’, ‘Canonical pathways’, ‘Transcription factor targets’ and/or KEGG pathway gene sets [22, 23]. Significantly enriched gene sets in the gsea experiment were table-summarized as previously described [24].

Real-time quantitative PCR

Real-time experiments were carried out as described previously [15]. Primers are listed in Table S2. For MKYSs, reverse transcription was performed directly on cells resuspended in 10 μL of water after an initial denaturation for 5 min at 70 °C. Statistical analyses were performed in prism (GraphPad, San Diego, CA, USA), with an unpaired t-test (Mann–Whitney). Otherwise, the stated results are mean expression relative to the PPIA gene or RNU6B small RNA of at least three samples ± standard error of the mean.

Results

Ex vivo generation and characterization of MKs at different developmental stages

CD34+ progenitor cells from fetal liver, cord blood and adult blood or bone marrow were induced by SCF and TPO to differentiate to the MK lineage (Fig. 1A). To study primitive hematopoiesis, we generated MKs from hESC lines. For this purpose, we used four cell lines (H1, H9, HUES 1, and HUES 7), following a previously described protocol [16], with slight modifications to avoid manual selection of colonies (Fig. 1B).

Figure 1.

Megakaryocyte (MK) differentiation and characterization. (A) CD34+ cells from fetal liver, cord blood or adults were differentiated into MKs until day 14. (B) Human embryonic stem cells (hESCs) were differentiated into MKs with a modified version of Takayama's protocol [16] until days 22–24. (C) Kinetics of surface marker expression during MK differentiation from hESCs (hESC-derived MKs [MKESs]), fetal liver-derived CD34+ cells (MKFLs), cord blood-derived CD34+ cells (MKCBs), or adult blood CD34+ cells (MKADs). Results are expressed as percentage of CD41high gated cells. (D) May–Grünwald–Giemsa-stained cytospins of CD41+/CD42+ cells from MKES, MKFL, MKCB or MKAD differentiation. (E) Flow cytometry ploidy distribution of MKESs, MKFLs, MKCBs, or MKADs. n is the calculated mean ploidy. (F) Flow cytometry analysis of cell surface MPL. The red line represents MKs as compared with the Ig isotype control (blue). (G) Flow cytometry analysis of cell surface P-selectin (CD62) after thrombin activation (red line) or before thrombin activation (blue line). (H) Percentage of proplatelet-forming MKs during ontogeny. Results are mean ± standard deviation from three replicates from at least two experiments. bFGF, basic fibroblast growth factor; SCF, stem cell factor; Thrb., thombin; TPO, thrombopoietin.

The kinetics of MK production were studied by use of an anti-CD41 antibody. In cultures of fetal liver, cord blood and adult CD34+ cells, the number of CD41+ cells increased from day 4 to reach a plateau between days 10 and 12, and these cells were detected until day 20. In hESC cultures, the picture was slightly more complex: CD41+ cells corresponding to CD41+/GPA+ early progenitors mainly destined for the erythroid lineage first peaked around day 7 [12, 25], and then a mixed population committed to MK differentiation (CD41high) or to macrophage differentiation (CD14+/CD15+/CD41low cells) peaked at days 10–15 (data not shown). To obtain a more enriched MK population, the CD41high cells were sorted at day 11 and further cultured on OP9 cells in the presence of TPO and SCF (Fig. 1B).

The kinetics of CD41high cell differentiation explored with a panel of differentiation markers (Fig. 1C) were similar at different developmental stages, with the exception of CD34 and CD43, whose expression was sustained specifically in embryonic stem cell cultures. In particular, the more mature MKs coexpressed CD41, CD42, CD9, and CD49e, but not CD34. Both morphologic analysis and ultrastructural studies (not shown) confirmed the MK maturation (Fig. 1D). However, in contrast to MKADs, MKESs, MKFLs and MKCBs were rarely large with polylobulated nuclei, but mostly small with round and indented nuclei correlating with their low ploidy level (Fig. 1E). All of these MKs expressed MPL at their cell surface (Fig. 1F), and could be activated as platelets by thrombin, which induced CD62 surface expression (Fig. 1G). Finally, all MK populations were able to form proplatelets, at a frequency ranging from 5% for MKFLs to 28% for MKADs (Fig. 1H).

Taken together, these observations demonstrate that human MKs share common cellular characteristics, and that MKESs resemble MKFLs, as shown by a reduced propensity to undergo polyploidy and to shed platelets as compared with adult MKs.

Gene expression profiling of MKESs, MKFLs, MKCBs, and MKADs

MKESs showed a mature-like pattern of expression from day 10 of differentiation. To precisely determine at which day they were really mature, in order to perform a comparison with the other MKs, we first compared day 14 and day 18 MKESs (defined as cells expressing high level of CD41 and CD42 markers), and found that the day 14 MKES population still corresponded to an immature cell population expressing genes associated with immaturity, such as MYH10, or lacking expression of genes, such as F2R or the von Willebrand factor gene (VWF), that are associated with mature MKs (Data S1; Fig. S1). Thus, to decipher the developmental changes at the molecular level between mature MKs at different developmental stages, we compared the gene expression of MKs derived from four hESC lines (MKES1 to MKES4) at day 18, three fetal liver samples (MKFL1 to MKFL3), three cord blood samples (MKCB1 to MKCB3), and three adult blood samples (MKAD1 to MKAD3). All hybridizations were performed twice, each time in duplicate (dual color), with one adult sample (MKAD1) as a reference, thus providing two technical replicates for each sample, and 28 technical replicates of the MKAD1 profile. A hierarchical clustering algorithm and an r-based principal component analysis were applied, and revealed that MK samples segregated according to their developmental stage (Fig. 2A,B). MKESs from the different hESC lines were closely related one to another and also to MKFLs. There was a maximal difference between the transcription profiles of MKESs and MKADs, and intermediate differences between MKCBs and MKFLs. To determine whether these differences had functional consequences, we performed a class comparison analysis, defining each developmental stage as a statistical class. In a first approach, we analyzed the most highly expressed genes of each class and the genes specifically overexpressed in each class as compared with the others. The GO analysis of these data revealed that all samples share common characteristics, and have moderate developmental differences.

Figure 2.

Genome profiling characterization of MKs. (A, B) Non-supervised hierarchical clustering with sample cluster P-values, using dchip software (A) and principal component analysis analysis of all MK samples (B). Blue spheres represent human embryonic stem cell MK (MKES) samples, red spheres fetal liver-derived MK MKFL samples, green spheres cord blood-derived MK (MKCB) samples, and orange spheres adult MK (MKAD) samples. (C) Hierarchical representation of the 2021 genes obtained from pairwise comparisons between developmental stages. (D) Numbers of genes classified according to their differential expression levels between MKESs and MKFLs (left panel), MKFLs and MKCBs (middle panel), and MKCBs and MKADs (right panel). See Fig. S2A for additional comparisons. (E) Real-time quantitative PCR validation in MKYSs (dotted bar) of genes differentially expressed between MKESs and MKFLs, and MKCBs and MKADs. Results are mean expression relative to the PPIA gene from two samples for MKYSs. (F) gsea analysis of pairwise comparisons between developmental stages to identify GO biological processes (left panel), Broad Institute canonical pathways (middle panel) or Broad Institute transcription factor targets (right panel) enriched (red) or depleted (blue). Only significant enrichments are shown (< 0.05; false discovery rate < 0.25). See Table S4A–C for other comparisons. BMP, bone morphogenetic protein; ECM, extracellular matrix; IL, interleukin; VEGFR1, vascular endothelial growth factor receptor 1.

To further analyze the differences between different developmental stages, we performed a pairwise comparison between each class, with a classic two-fold change and P < 0.001 as a cut-off; this revealed 2021 differentially expressed genes that allowed hierarchical clustering of the developmental stages (Fig. 2C). However, most of these genes were highly expressed only in MKESs, and showed progressivily decreased expression in the subsequent stages (MKFL > MKCB > MKAD). Moreover, a majority of these genes showed only two-fold to four-fold expression changes between two contiguous developmental stages (Fig. 2D; Fig. S2A).

To verify that MKs derived from hESC lines could be assimilated as primitive human MKs, we performed real-time validation of several genes that are differentially expressed between MKESs and MKFLs in MKs derived from the human embryo yolk sac (Fig. 2E). These experiments confirmed that MKESs and MKYSs are similar at the transcriptional level, and strengthen the use of MKESs as a surrogate for primitive MKs.

gsea analysis was applied between all populations (Fig. 2F; Table S3A–C). With the use of GO categories, one striking result was the enrichment in cell cycle processes (e.g. ‘Cell cycle’, ‘Mitosis’, ‘DNA replication’) in the MKAD class. This was related to enrichment in E2F transcription factor (TF) target genes and in pathways such as Aurora A/B or PLK1, involved in ploidization (Table S4). Other genes implicated in cell cycle regulation, such as those encoding MCM proteins, CENP proteins, or cyclin A2, were enriched in MKCBs and MKADs as compared with MKESs and MKFLs. Enrichment in genes implicated in protein production (those encoding nucleoporins or small nuclear ribonucleoproteins) and lipid metabolism, such as those encoding ACACA and GPAT2, was also observed, especially in the MKES vs. MKFL and MKES vs. MKAD comparisons (Table S4). Finally, there was relevant enrichment in ‘Factors involved in MK development and platelet production’, such as CXCR4, CD36, VWF and ALOX12 in the later developmental stages (Table S4). In contrast, there were no changes in genes involved in mitochondrial function. Taken together, these data further support the hypothesis that, during ontogeny, MKs undergo a progressive process of specification towards highly efficient production of functional platelets.

In contrast, several other biological functions were enriched in the embryonic and fetal developmental stages, including regulation of angiogenesis (Fig. 2F), mainly the transforming growth factor-β receptor, VEGF receptor 1 and bone morphogenetic protein signaling pathways, extracellular stimuli, growth factor response, activation of integrin pathways, and cytoskeleton organization (Table S4). Finally, there was also enrichment in potential targets of SRF or GATA1 that have been implicated in angiogenesis and hematopoiesis in mice and zebrafish. Consistent with a previous report in mice, we found enrichment of MYB potential transcription target genes in the later developmental stages [5] (Fig. S2B).

Ontogeny-driven genes

To identify the genes that were regulated all along the different developmental stages, we next filtered gene expression by increasing or decreasing intensity throughout ontogeny, and identified 695 genes fitting these criteria, with 442 showing decreased expression and only 253 showing increased expression (Fig. 3A,B; Table S5).

Figure 3.

Ontogeny-driven genes. (A, E) Hierarchical representation of ontogeny-driven genes (A) and transcription factors (E). (B) Numbers of differentially expressed genes conditional on changes in expression levels (red bar, increased; green bar, decreased) between human embryonic stem cell megakaryocytes (MKESs) and adult megakaryocytes (MKADs). (C, D, F) Real-time quantitative PCR validation of ontogeny-driven genes. MKCB, cord blood-derived megakaryocyte; MKFL, fetal liver-derived megakaryocyte; VWF, von Willebrand factor.

Several genes that were upregulated during ontogeny were directly involved in megakaryopoiesis, platelet development, or platelet function and aggregation, such as ITGA6 (CD49f), CXCR4, ADCY3, DKK1, GNA15, VWF, CLU, SLC6A4 (SERT), and MAOB (Fig. 3C).

A different pattern of expression was found for downregulated genes. A large fraction of them showed a drastic decrease in expression between MKESs and MKADs. These genes were related to the extracellular matrix, membrane components, and cytoskeleton (COL1A2, COL1A1, LAMB1, KDR, APBB1IP, ALAS2, HMOX1, CD24, AKR1C1, and AVPR1A) (Fig. 3D).

The differential expression of TFs was sufficient to discriminate between early and late developmental stages, but failed to discriminate between MKCBs and MKADs (Fig. 3E,F). Increased expression of TF genes such as STAT5A, ETS2 and DLX1 supported the hypothesis that MKs become more specialized in platelet production [26, 27]. The expression of genes encoding homeodomain-containing TFs, such as PITX2, HOPX, MEIS2 and PBX, also changed during ontogeny. Several Notch-related TF genes, such as HEY1, HES4, and MYCN, were expressed in the early stages of ontogeny and showed decreased expression at later stages. Expression in the first stages of ontogeny was also particularly marked for LIN28B (Fig. 3F), previously described as a key marker of the fetal stage [28, 29]. MYC has been already reported to be a key TF in the MK lineage [30], but the expression of its homolog MYCN, particularly in MKESs and MKFLs, was unexpected, although it has been implicated in the regulation of HSCs and progenitor cells in the fetus [31].

To obtain a global view of gene expression throughout the course of MK ontogeny, we performed an miRnome analysis in MKESs and MKADs. Global analysis of differentially expressed miRNAs with IPA Ingenuity revealed very low levels of ontogenic change, particularly in specific canonical pathways, or changes in any biological functions. Thus, we carefully examined putative relationships between ontogeny-driven genes and differentially expressed miRNAs. We observed that 21 miRNAs were significantly more expressed in MKESs than in MKADs (Fig. 4A; Table S6). Notably, miR145, which has been implicated in MYC downregulation [32], was inversely correlated with MYC expression throughout ontogeny (Fig. 4B). Two other miRNAs, miR224 and miR9, that are predicted to regulate CXCR4, were significantly more highly expressed in MKESs than in MKADs (Fig. 4C). This could contribute to the significant increase in CXCR4 mRNA during ontogeny (Fig. 3C). In MKADs, 11 miRNAs were significantly more expressed than in MKESs (Fig. 4A); in particular, miR181a has been previously implicated in MK differentiation and directly induces LIN28 downregulation [33]. This higher expression of miR181a in MKADs (Fig. 4D) was correlated with higher expression of LIN28B in MKESs. Moreover, several let-7 miRNAs (let-7a, let-7f, and let-7g), which are known as direct targets of LIN28B, were upregulated in MKADs as compared with MKESs (Fig. 4E). Together, these results indicate the fine regulation of several major TFs and regulators during MK ontogeny.

Figure 4.

Ontogeny-driven microRNAs (miRNAs). (A) Hierarchical representation of the 32 miRNAs differentially expressed between human embryonic stem cell megakaryocytes (MKESs) and adult megakaryocytes (MKADs). (B–E) Quantification of miR145, miR9, miR224, miR181A, let-7a, let-7f and let7-g expression in MKESs and MKADs, based on miRnome fluorescence intensity (upper panels) and in all developmental stages, using real-time quantitative PCR (RT-qPCR) (lower panels). (B) RT-qPCR validation of MYC expression. MKCB, cord blood-derived megakaryocyte; MKFL, fetal liver-derived megakaryocyte.

Discussion

This study presents the first global comprehensive analysis of the transcriptome of human MKs during development.

Gene expression profiling of MKs derived from four different hESC lines identified limited differences, indicating that they represent a truly specific entity. However, MKESs were more closely related to MKFLs than to the two other developmental stages. Previously, only MKESs and MKFLs have been compared in the mouse, and differences were found in the expression of different proteins and TFs [34]. We provide strong evidence that MKES are close to primitive MKs. First, erythroblasts in the same cultures expressed embryonic and fetal globin chains [12]; and second, several genes differentially expressed between MKESs and MKFLs were similarly expressed in MKYSs.

As previously demonstrated for erythroblasts and HSCs [24, 29], the changes in expression levels between genes differentially expressed between two contiguous stages of ontogeny were usually modest (less than two-fold). This suggests that, whatever their stage of ontogeny, MKs exhibit a common signature and that fine-tuning occurs during development. In favor of this hypothesis, of the 695 genes that were continuously regulated during MK ontogeny, the 253 upregulated genes mainly affect MK polyploidization, the cytoskeleton, proplatelet formation, and platelet functions, indicating a development-associated increase in adult MK-specific hemostatic functions, which might be less important during fetal life, as attested by p45NF-E2 knockout mice [35]. Conversely, platelet–endothelial interactions are required around day 13.5 for separation of the blood from the lymphatic vasculature [36], and for the brain vasculature to develop correctly. No marked changes in MYH10 expression were found in mature MKs, whatever the ontogenic stage. This strongly suggests that sustained expression of MYH10 during the entire course of MK differentiation is not the mechanism explaining the low ploidy of embryonic/fetal MKs. However, from these experiments, we cannot exclude the possibility that MYH10 may be expressed at higher levels in immature embryonic/fetal MKs. A set of genes involved in the interaction of the cell with its environment are upregulated in MKESs and MKFLs, and may contribute to MK adaptation to the changing environment (yolk sac, fetal liver, and bone marrow) [37]. It is noteworthy that CXCR4 is developmentally regulated by an off–on mechanism. CXCR4 expression was very low in MKESs and MKYSs, and its expression was greatly increased in MKADs. The CXCR4–SDF1 pathway is required for platelet production in the adult, by allowing MK localization in the bone marrow endothelial niche and proplatelet release into the blood flow [38]. In early MKs, miR9 and miR224, which both target CXCR4 transcripts and undergo developmental regulation, could be responsible for this silencing [39].

During MK ontogeny, LIN28B is specifically expressed in MKESs and MKFLs. Fetus-specific expression of LIN28B has also been demonstrated in erythroid and T-cell lineages [29, 40]. Its ectopic expression in adult hematopoietic progenitors leads to fetal lymphopoiesis [40]. These changes in LIN28B expression in MKs are probably attributable to miR181a, which is inversely expressed during development. Moreover, LIN28B has been implicated in a regulatory network with MYC factors [41], and this could be related to our observation of specific MYCN expression during the early stages of ontogeny. Fine-tuning of MYC expression, including downregulation and then upregulation, is essential in MK amplification, polyploidization, and proplatelet formation from induced pluripotent stem cells [42].

With the exception of MYB, the expression levels of major TFs of the MK lineage were not significantly changed during MK ontogeny, in accordance with observations made in HSCs and erythroblasts [24, 29]. Nevertheless, we suggest that GATA1, RUNX1 and SRF have increased gene targets at early developmental stages (MKESs and MKFLs) as compared with MKADs, suggesting a more important transcriptional activity. The increased activity of SRF could account for differences between embryonic/fetal and adult MK stages [43].

It is noteworthy that two subgroups of pediatric AMKL, which arise during fetal development, involve oncogenes implicated in these pathways, namely GATA1 genes in Down syndrome, and MAL in OTTMAL fusion-associated AMKL [8, 44, 45]. Similarly, the enrichment in TFs of the Notch pathway in embryonic/fetal MK stages correlates with the central role of Notch pathway deregulation by OTT in the pathogenesis of OTTMAL AMKL [46]. These observations further support the link between pediatric AMKL and the regulation of embryonic/fetal megakaryopoiesis.

In conclusion, this study clearly reveals important differences between MKESs and the other developmental stages, and it demonstrates that development is associated with MK maturation leading to increased homeostatic functions that are vital for adult life.

Addendum

O. Bluteau, T. Langlois, P. Rivera-Munoz, F. Favale, P. Rameau, G. Meurice, P. Dessen, T. Mercher, and W. Vainchenker: participated in designing and/or performing the research; O. Bluteau, T. Langlois, G. Meurice, P. Dessen, and W. Vainchenker: controlled and analyzed data; O. Bluteau and W. Vainchenker: wrote the manuscript; T. Langlois, P. Dessen, T. Mercher, E. Solary, H. Raslova, and N. Debili: participated in writing the manuscript.

Acknowledgements

The authors thank F. Wendling for critical reading of the manuscript, Y. Lecluse for technical support in cytometry, T. Robert and C. Oréar for microarray hybridization, and S. Kocielny for helpful discussions on statistical methodology.

This work was supported by grants from the Agence Nationale de la recherche (ANR-09-BLAN-0275), the Association pour la recherche sur le cancer (ARC 3193; W. Vainchenker), and Labex GR-Ex, which is funded by the program ‘Investissements d'avenir’. O. Bluteau, P. Rivera-Munoz and T. Langlois were supported by a postdoctoral fellowship from ANR.

Disclosure of conflict of interests

The author state that they have no conflict of interest.

Ancillary