Megakaryocytes promote hepatoepithelial liver cell development in E11.5 mouse embryos by cell-to-cell contact and by vascular endothelial growth factor A signaling


  • Potential conflict of interest: Nothing to report.

  • This work was supported by grants from the Ministerio de Ciencia e Innovación (MICINN; grant nos. SAF2007-65265 and SAF2009-12596), the Instituto de Salud Carlos III (grant no.: ISCIII08/1374), and the Comunidad Autónoma de Madrid (grant no.: SAL-0304-2006). N.S. and I.C. are recipients of fellowships from the Centro de Biología Molecular Severo Ochoa (CBMSO) and the MICINN, respectively. The CBMSO receives institutional funding from the Fundación Ramón Areces.


In the mouse embryo, hematopoietic progenitor cells migrate to the fetal liver (FL) between gestational days (E) 9.5 and 10.5, where they rapidly expand to form the main fetal reservoir of hematopoietic cells. The embryonic megakaryocyte progenitors (MKPs) in the E11.5 FL were identified as CD49fHCD41H (and c-KitDKDR+CD42+CD9++CD31+) cells, expressing several hepato-specific proteins. Unlike adult bone marrow megakaryocytes (MKs), embryonic MKPs were CD45 and represent an abundant population in the FL. The CD49fHCD41H MKPs purified by cytometry differentiated in vitro to produce proplatelets, independent of thrombopoietin stimulation, and they responded to stimulation with adenosine diphosphate, thrombin, and the PAR4 thrombin receptor-activating peptide. Moreover, after removing CD49fHCD41H MKPs from purified E11.5 FL hepatoepithelial-enriched cell preparations (c-KitDCD45Ter119), the remaining CD49fD cells neither differentiated nor survived in vitro. Indeed, direct cell-to-cell contact between the CD49fHCD41H and CD49fD populations was required to promote the hepatocyte differentiation of CD49fD cells. The addition of vascular endothelial growth factor A (VEGF-A) and medium conditioned by E11.5 CD49fHCD41H MKPs produced a partial effect on CD49fD cells, inducing the formation of hepatoepithelial layers. This effect was abolished by anti-VEGF-A antibodies. Together, these findings strongly suggest that CD49fHCD41H MKPs are fundamental to promote FL development, as proposed in adult liver regeneration. Conclusion: The cells of the MK lineage present in the developing mouse embryo liver promote the growth of hepatoepithelial cells in vitro through VEGF-A signaling and may play a role in liver development in vivo. (HEPATOLOGY 2012;56:1934–1945)

After gastrulation, genetic prepatterns are established in discrete areas of the embryo related to cell-lineage specification, cell differentiation, and morphogenesis. Hematopoiesis occurs in two phases in the embryo (primitive and definitive). Primitive hematopoiesis involves embryonic erythrocytes and myeloid cells, commencing in the yolk sac (YS) and proceeding as a self-limiting process throughout gestation. By contrast, definitive lymphohematopoiesis begins in the YS and, in an autonomous manner, in the para-aortic splanchnopleura/aorta-gonads-mesonephros (P-Sp/AGM) niche, which later becomes the source of all lymphoid and hematopoietic cell lineages.1-3

Megakaryocytes (MKs) are a particular blood cell type that share common features with hematopoietic stem cells (HSCs). In the adult, CD45+CD9+ CD41++ MKs are found primarily in the bone marrow (BM) as a scattered polyploid population of large cells. These MKs are responsible for the production of platelets, subcellular fragments involved in coagulation and the regulation of angiogenesis.4 In the mouse embryo, clonogenic bipotential megakaryocyte/erythroid progenitors (MEPs) appear in the YS at embryonic days 7.25 (E7.25) and E9.5, participating in primitive and definitive megakaryopoiesis, respectively.5, 6 At E10.5, large, immature reticulated platelets have been found in the bloodstream, and CD45CD41H cells can be observed in vascular hematopoietic clusters.5, 7 With the discovery of thrombopoietin (TPO), MKs could be cultured, and platelets were generated in vitro from mature MKs, isolated by density purification, that produce long, pseudopodial cytoplasmic processes (i.e., proplatelets).8

After E10.5, the fetal liver (FL) represents the central compartment of hematopoiesis during gestation and receives extrinsic progenitors derived from the primary hematopoietic niches (YS, P-Sp/AGM, and the placenta), although it also harbors endodermal progenitors derived from the gut evagination that later give rise to the adult liver parenchyma. It has been suggested that cross-talk between the hematopoietic and hepatoepithelial compartments plays a central role in liver development.9 Bipotential hepatocyte and cholangiocyte progenitors (HeP) have been identified in the mouse embryo.10-12 Indeed, we identified an embryonic HeP population that was negative for hematopoietic markers (CD45Ter119), but that weakly expressed the stem cell factor receptor (c-KitD) and which could be separated into two subpopulations based on the level of α6 integrin chain expression (CD49f). The amount of CD49f expression in HeP remains unclear, with some studies describing HeP as CD49f negative11 and others describing postnatal liver progenitor cells as CD49fH.13 In the present study, we demonstrate that, at E11.5, the CD49fH subpopulation of c-KitD cells are functional MK precursors (MKPs) that are CD41HCD42a,b,c+CD9++. Furthermore, unlike the precursors from the adult BM, this population lacks the conventional hematopoietic tracer (CD45) and these cells express vascular endothelial growth factor A (VEGF-A). When cultured in vitro in the absence of TPO, these embryonic MKPs produce proplatelets, which are also clearly evident directly among the cells isolated from FL. Finally, we show that the CD49fHCD41H MKPs present in the FL of E11.5 embryos establish numerous contacts with albumin (ALB)+ cells in vivo and stimulate the development of CD49fD HeP in vitro in response to direct cellular contacts and VEGF-A.


AAT, α1-antitrypsin; Abs, antibodies; ADP, adenosine diphosphate; AGM, aorta-gonads-mesonephros; ALB, albumin; AFP, alpha-fetoprotein; APC, allophycocyanin; BM, bone marrow; cDNA, complementary DNA; Col I, collagen I; CytoB, cytochalasin B; E, gestational day; FACS, fluorescence-activated cell sorting; FITC, fluorescein isothiocyanate; FL, fetal liver; FSC, forward-scattered light; GαS, G-protein subunit αs; GLUT2, glucose transporter type 2; GPIbα, glycoprotein Ibα; HeP, hepatocyte and cholangiocyte progenitor; HGF, hepatocyte growth factor; HNF, hepatocyte nuclear factor; HSCs, hematopoietic stem cells; IF, immunofluorescence; iMKs, immature megakaryocytes; ISO, isotype-matched Abs; KDR, kinase domain region; MEPs, megakaryocyte/erythroid progenitors; MK, megakaryocyte; MKP, megakaryocyte progenitor; NES, nestin; PBLs, peripheral blood lymphocytes; PCR, polymerase chain reaction; PE, phycoerythrin; P-Sp, para-aortic splanchnopleura; SEM, standard error of the mean; TPO, thrombopoietin; TTR, transthyretin; VEGF-A, vascular endothelial growth factor A; VEGFR2, VEGF receptor 2; VIM, vimentin; VWF, von Willebrand factor; VWFR, VWF/thrombin receptor; YS, yolk sac.

Materials and Methods

Embryo Microsurgery and Cell Suspensions.

BALB/c and C57BL/6 mice were maintained at the animal facility of the Instituto de Salud Carlos III (Madrid, Spain). Mice were mated overnight, and the day the vaginal plug was detected was considered day 0.5 of gestation (E0.5). Mice were sacrificed, exsanguinated to collect peripheral blood lymphocytes (PBLs), and the desired tissues (FL and AGM) were obtained as described in the Supporting Methods. All animal studies were approved by the institutional review boards of the Instituto de Salud Carlos III and the Centro de Biología Molecular Severo Ochoa (Madrid, Spain).

Flow Cytometry and Cell Purification.

Cells were treated with Fc-Block (BD Biosciences, San Diego, CA) and 10% normal mouse serum in phosphate-buffered saline before incubation (4°C, 30 minutes) with fluorescein isothiocyanate (FITC), biotin-, phycoerythrin (PE)-, or allophycocyanin (APC)-conjugated antibodies (Abs) (as indicated in Supporting Table 1). Cell debris and dead cells were excluded by light-scattering parameters and propidium iodide staining, and cell suspensions were analyzed in a FACSCalibur with the CellQuest (BD Biosciences) and FlowJo (Tree Star Inc., Stanford University, Stanford, CA) software packages. Cells were purified under sterile conditions by fluorescence-activated cell sorting (FACS) on a FACSAria (BD Biosciences), and the purity of the cells recovered was over 98%. For cell-culture experiments, staining with Abs directed against c-Kit/FITC, CD49f/PE, CD45/biotin and Ter-119/biotin (visualized with streptavidin/APC) allowed c-KitDCD45Ter119, c-KitDCD45Ter119CD49fH, and c-KitDCD45 Ter119CD49fD cells to be purified (forthwith referred to as c-KitDCD45, CD49fH, and CD49fD, respectively). Because purified CD49fH and CD49fD cells were CD41HCD42c+CD45 and CD41 CD42cCD45, respectively, for some experiments, CD49fHCD41H and CD49fD cells were purified after staining with Abs directed against CD41/FITC, CD45/APC, and CD49f/PE.

Platelet and MKP Functional Assays.

Cells were collected, activated with adenosine diphosphate (ADP), thrombin, and the PAR4 thrombin receptor-activating peptide, and analyzed by flow cytometry (as indicated in the Supporting Methods).

Polymerase Chain Reaction.

RNA was extracted and oligo(dT)-primed complementary DNA (cDNA) was prepared as previously described.14 Polymerase chain reaction (PCR) amplification was performed with the primers and conditions indicated in Supporting Table 2 and Supporting Methods. For quantification of ALB and kinase domain region (KDR) expression, quantitative real-time PCR was performed as previously described.15 The relative amount of specific cDNA on each sample was determined by the 2−ΔΔCt method using G-protein subunit αs (GαS) expression as an internal control.

Cell Culture.

Purified E11.5 FL cells (1-2 × 105 cells/cm2) were cultured for 1-7 days under the general conditions detailed in the Supporting Methods. In some experiments, the following soluble factors were added: murine TPO (50 ng/mL; PeproTech, London, UK); VEGF-A (1-10 ng/mL; PeproTech); serotonin (1 μM; Sigma-Aldrich, St. Louis, MO); and cytochalasin B (CytoB; 20 μM; Sigma-Aldrich). When indicated, purified Leaf antimouse VEGF-A or Leaf isotype-matched Abs (ISO; clones 2G11-2A05 and RTK2758, respectively; BioLegend, San Diego, CA) were added to cultures. For coculture assays, CD49fD purified cells were plated in the upper transwell chamber of Col I/III-coated membranes (0.4-μm pore size) in 24-well plates (Costar, Cambridge, MA). CD49fH purified cells were plated either in the lower chamber or in combination with CD49fD cells in the upper chamber. After 7 days, cells in the upper chambers were collected for RNA extraction. Representative images from cultures were captured on a Leica DMI3000B microscope equipped with a DFC420 camera (Leica, Wetzlar, Germany). For scanning electron microscopy, cultures were treated as indicated in the Supporting Methods.


The slides that were stained contained the following: (1) cytospin preparations of sorted or unpurified FL cells prepared by centrifugation in a Cytospin-4 (65 G, 5 minutes; Shandon Southern Products, San Jose, CA); (2) cells cultured on Col I-coated slide chambers; and (3) E11.5 frozen tissue sections (7 μm thick). Samples were treated and stained as indicated in the Supporting Methods. The resulting images were processed using the ImageJ software (v1.43; National Institutes of Health, Bethesda, MD).

Calculation of Contact Frequencies Per Cell-Surface Area.

Contact frequency of CD49fHCD41H MKPs per cell surface area was calculated as described previously,16 dividing the number of contacts observed between MKPs (as CD41H cells) and ALB+ cells or MKPs, and with c-Kit+ cells, by the total surface area of these populations. Confocal immunofluorescence (IF) images were used to measure the corresponding cell radius and to determine frequencies in each population and cell contacts observed between them. Total surface area of each population is the product of the mean surface area of single cells by the total cell numbers of this population. The number of FL cells counted at E11.5 was 95,690 ± 7,110/organ (n = 10).

Statistical Analysis.

All data are presented as the means ± standard error of the mean (SEM) that were calculated with GraphPad Prism 4.0 software (GraphPad Software, Inc., La Jolla, CA), and the unpaired t test and the chi-square test were applied.


A Heterogeneous c-KitDCD45 HeP Population Is Present in the FL at E11.5.

CD49f expression in the c-KitDCD45 cell subpopulation of E11.5 FL was characterized by performing a detailed flow-cytometry phenotypic study. Expression of CD45 and either the VEGF receptor 2 (VEGFR2; recognized by the KDR marker) or the integrin αIIb chain (GPIIb/CD41) was quantified in electronically gated FL c-KitDCD49fH and c-KitDCD49fD cells (Fig. 1A-C). A large number of CD49fH cells were either CD45KDR+/CD41++ or CD45++KDR/CD41 (CD49fHCD41H and CD49fHCD45H, respectively), whereas a small proportion were CD45+KDR+CD41+. By contrast, most CD49fD cells did not express CD45, CD41, or KDR. The panhematopoietic marker (CD45) labels myeloid-derived cells in the early embryo, and indeed the majority of CD49fHCD45H cells were positive for CD11b/Mac1 (Fig. 1D). High levels of CD41 expression in adult BM is characteristic of MKs, whereas low levels are typical of HSCs. Analysis of the c-Kit receptor, the tetraspanin molecule (CD9), and of the CD42a, CD42b, and CD42c chains of the von Willebrand factor (VWF)/thrombin receptor (VWFR) revealed that CD49fHCD41H cells were c-KitD, CD9++, and VWFR+, a phenotype more characteristic of MKPs and not HSCs. These cells were large (mean forward-scattered light [FSC] intensity, 545 ± 24; n = 5), did not correspond to cellular fragments or platelets, and were abundant, representing up to 8 × 103 ± 282 (n = 10) of viable cells in the FL at E11.5. This initial analysis indicated that the E11.5 FL c-KitDCD49fH cell subset identified had a surface phenotype compatible with CD41H (and CD9+VWFR+) MK cells. In functional analyses performed on FL cell suspensions stimulated with ADP, thrombin, and the PAR4 peptide, CD41H cells up-regulated the active form of the CD41/CD61 fibrinogen receptor and fibrinogen binding as well as inducing actin polymerization (Table 1).

Figure 1.

Phenotypic characterization of CD49f cells present in the FL of E11.5 mouse embryos. FL cells stained with the indicated Abs were analyzed by flow cytometry, displaying the results in (A-E) as representative dot plots or histograms. The background level from isotype-matched irrelevant Abs is indicated by overlaid dotted lines in the histograms. Fluorescence scales are logarithmic, and the FSC scale is linear. (A) Gates inside the dot plot showing anti-CD49f and anti-c-Kit staining identify the CD49fH and CD49fD subpopulations among the c-KitD cells (n = 15). (B and C) Staining with anti-CD49f/CD45/KDR or anti-CD49f/CD45/CD41. Gates in the dot plot in (B) define the CD49fH (C, upper plots) and CD49fD (C, bottom plots) subpopulations in which the expression of CD45 and KDR or of CD45 and CD41 were analyzed. The numbers represent the percentage of cells in each quadrant (mean ± SEM; n = 5). The gates in the upper right dot plot in (C) label the CD45HCD41 and CD45CD41H populations in which Mac1, c-Kit, CD9, and CD42a, CD42b, and CD42c expression was analyzed in (D), as indicated in the histograms highlighted by arrows (n = 3). (E) Expression of integrin and receptor molecules in CD45CD41H cells. FL cells were stained with anti-CD45/CD41 and with the Abs indicated below each histogram, which display the signal obtained for each specific Ab in the gated CD45CD41H population (upper dot plot), whose percentages are shown in the right (n = 4). (F) Quantification of these data is shown as the percentage of cells positive for each Ab (bar graphs). Mean fluorescence intensity is indicated by the numbers above each column (mean ± SEM; n = 4).

Table 1. Functional Assays in Resting and Activated MKP and Platelets
Exp. no.SampleNoneADP ApyADPPAR4Thrombin
  1. a

    Values represent mean fluorescent intensity.

  2. Abbreviations: exp. no., experiment number; Apy, apyrase.

Surface expression of active conformation of CD41/CD61
 1E11.5 MKP366a392483  
 1Adult platelets1723.247.2  
 2E11.5 MKP790 821862 
 2E11.5 platelets187 232279 
 2Adult platelets60.1 99.3200 
 3E14.5 MKP251 356323 
 3E14.5 platelets56.5 147122 
 3Adult platelets48.9 61.8106 
Fibrinogen binding to CD41/CD61-positive cells
 4E11.5 MKP677 1,0967811,070
 4Adult platelets82.5 173377242
Actin polymerization in CD41-positive cells
 5E11.5 MKP775 929932933
 5Adult platelets463 648531593
 6E11.5 MKP692 745723739
 6Adult platelets659 684674711

When the expression of other integrin and receptor molecules was analyzed (Fig. 1E,F), most of these electronically gated CD49fHCD41H cells expressed different levels of α4 (CD49d), α5 (CD49e), αV (CD51), αL (LFA1/CD11a), β1 (CD29), β2 (CD18), and β3 (CD61) chains as well as the endothelial marker CD31 and the intercellular adhesion molecule-1 (ICAM1, CD54). Approximately half of the CD49fHCD41H cells were positive for the integrin α2 chain (CD49b), and these cells had no αM (CD11b/Mac1) and β4 (CD104) chains nor the vascular cell adhesion molecule-1 (VCAM1, CD106), receptors for fms-related tyrosine kinase 3 (Flt3), or the lymphohematopoietic marker AA4.1. Taken together, these findings demonstrate the presence of significant numbers of cells with a surface phenotype and a functional behavior characteristic of MKPs in the E11.5 FL. These cells display several integrin receptors and are easily identified by flow cytometry as CD49fHCD41H (and CD9++CD42c+) cells.

E11.5 CD49fHCD41H Cells Spontaneously Develop Proplatelets In Vitro.

We used flow cytometry to purify and subsequently culture E11.5 CD49fHCD41H cells to analyze their differentiation in vitro. Cells with proplatelets were visible after 24 hours in culture in the absence of TPO. These cells had large cytoplasmic pseudopodia (mainly unbranched, with bulges along the shaft and a swelling at the tip), expressed CD41 and CD42c, and became prominent after 48 hours (Fig. 2A,B). In these conditions, few cells were attached to the surface of the plates, although the addition of TPO to cultures resulted in an increase in cell adherence (Fig. 2C) without affecting the number of MKs with proplatelets. On Col I-coated plates, attachment occurred earlier and the number of proplatelet-bearing MKs was higher. Staining with anti-CD41 and phalloidin after 48 hours in culture revealed different patterns of expression (Fig. 2D and Supporting Fig. 1), with some cells exhibiting a punctuate distribution of CD41, which colocalized with F-actin in podosome-like structures in adherent cells, and with F-actin clusters in the swellings along the shaft membrane of proplatelets. By contrast, in some cells, there were strong cytoplasmic accumulations of CD41, whereas others expressed CD41 primarily in microvilli and in membranes (indicative of the demarcation membrane system). When CytoB was added to recently seeded cultures, the formation of proplatelets was inhibited without affecting cell viability, and CD41 accumulated in the cytoplasm independently of F-actin (Fig. 2E). In conclusion, these in vitro data confirmed that embryo-derived CD49fHCD41H cells were MKPs capable of producing proplatelets in culture independently of TPO by an actin-dependent process.

Figure 2.

E11.5 FL CD49fHCD45 cells develop proplatelets. In vitro development of proplatelets from purified CD49fH cells (identified as in Fig. 1E) in short-term cultures grown on Col I-coated plates in the absence of TPO. Representative photomicrographs are shown after 48 hours of culture. (A) Bright-field photomicrograph. Arrowheads indicate swellings along the shaft. (B) IF photomicrograph showing a MK with proplatelets stained with Abs against CD41 (red) and CD42c (green). The merged image with the bright-field photomicrograph is shown. White arrow indicates the tip of a proplatelet. (C) Number of cells (as a percentage of the total) present in the cultures grown on uncoated plastic plates in the presence or absence of TPO, and on Col I-coated plates, after 24 and 48 hours. Unattached MKs with proplatelets (Prop-MK) or cells adhering to plates (ADH) were counted, and the data represent the mean ± SEM (n = 4). A total of 7 × 103 cells were counted in 125 image fields. **P < 0.01; ***P < 0.001 (unpaired t test). (D) Pattern of CD41 (green) and F-actin (red; phalloidin) expression; nuclei are counterstained with TOPRO-3 (blue). White arrows indicate podosome-like structures; arrowheads indicate swellings. (E) Effect of adding CytoB for 48 hours (n = 3).

CD49fHCD41H MKPs From E11.5 FL Display Hepatoepithelial and Endothelial Proteins.

Purified embryonic CD49fHCD41H MKPs exhibited a characteristic, punctuate VWF expression pattern in the cytoplasm (Fig. 3A) and were positive for ALB and nestin (NES; an intermediate filament expressed by endothelial and neural stem cells; Fig. 3B and Supporting Fig. 2). By contrast, CD49fD cells were ALB++ and were negative for NES. Isolated CD49fHCD41H MKPs were binucleated (and, less frequently, multinucleated) cells, some of which contained cytoplasmic protrusions, even after the mechanical stress produced by the FACS procedure (Fig. 3D). These proplatelets were more clearly observed when slides from unpurified E11.5 FL cells stained for CD41 were overexposed (Fig. 3E), indicating that fully developed proplatelets were not merely an in vitro differentiation product, but that they also existed in the E11.5 FL in vivo. The proplatelet-bearing CD41H cells present in unpurified FL were also ALB+ (Fig. 3F and Supporting Fig. 2).

Figure 3.

Ex vivo characterization of FL MKPs at E11.5. (A) Representative photomicrograph showing IF staining (red) with anti-VWF of a CD49fHCD41H MKP preparation (n = 3). (B and C) Photomicrographs of the immunodetection of ALB in purified CD49fHCD41H MKPs and CD49fD cells. The specific signal is shown in green; 91.4% ± 3.4% of the purified CD49fHCD41H cells were ALB+ (n = 350 cells from seven different experiments, 42 fields in total). (D) Representative bright-field photomicrograph from the hematoxylin and eosin staining of purified CD49fHCD41H MKPs (n = 3). Black arrows indicate the cytoplasmic pseudopodial extensions. (E) IF staining (red) of a total FL sample with anti-CD41. White arrows indicate proplatelets. (F) Dual IF staining of a total FL sample with anti-CD41 Ab (red) and anti-ALB (green). Photomicrographs show the signal of merged channels. White arrows are as in (E). For IF photomicrographs, nuclei were counterstained with TOPRO-3 (blue).

To determine whether these expression patterns were the result of NES and ALB synthesis by FL MKPs, we performed PCR analyses on total FL and YS cells, purified CD49fHCD41H MKPs and CD49fD cell populations from E11.5 FL, and adult tissues, including immature c-Kit+LinCD9+CD41+ MK (iMKs) isolated from BM.4 These analyses confirmed that VWF and the glycoprotein Ibα (GPIbα) chain of its receptor were expressed more strongly in CD49fHCD41H MKPs than in CD49fD cells. Moreover, CD49fH CD41H MKPs expressed VEGF-A and its receptor (KDR/VEGFR2), as well as NES, VIM, and several hepato-specific transcripts, such as ALB, alpha-fetoprotein (AFP), and transthyretin (TTR), although they did not express α1-antitrypsin (AAT) (Fig. 4A). IFs on tissue sections of E11.5 indicated that 60% ± 13% of CD41H cells express VEGF-A, and 27% ± 3% of these CD41HVEGF+ cells displayed the highest VEGF-A signal in FL (Fig. 4B and Supporting Fig. 3). There was a 20-fold increase in the expression of ALB transcripts in CD49fD cells when determined by quantitative real-time PCR (Fig. 4C). Expression of hepatoepithelial genes seemed to be specific to CD49fHCD41H MKPs of FL origin, because none were expressed in CD45CD41H MKPs isolated at E11.5 from other locations (such as the YS, AGM, and PBLs; data not shown) nor were they expressed in hematopoietic CD45HCD41 cells or in adult iMKs (Fig. 4D and Supporting Fig. 4). However, although dual-sorted purified CD45CD41H liver cells expressed ALB, AFP, and transthyretin, but not AAT, it cannot be fully ruled out that the CD49fH liver preparations were contaminated with a small number of CD49fD cells that could account for the hepato-specific signals.

Figure 4.

Gene expression in CD49fHCD41H cells. PCR analysis of megakaryocytic-, endothelial-, and hepatoepithelial-specific genes. (A) Panels show representative results from PCRs obtained for the following transcripts: VWF; GPIbα; KDR or VEGR2; VEGF-A; GαS; NES; VIM; TTR; AFP; ALB; AAT; and HGF. Four isoforms of VEGF-A transcripts were detected (left arrows). Preparations from at least three independent samples of total E11.5 FL and YS, adult liver (Li), kidney (Ki), and BM and from purified CD49fHCD41H MKP, CD49fD HeP, CD45+ cells, Ter119+ cells (erythrocytes, Ery), and BM megakaryocyte precursors (Linc-Kit++CD41++CD9++, iMK) were analyzed using genomic DNA and water as negative controls (C). GαS gene expression was used as control for messenger RNA content/sample. (B) Relative intensity of VEGF-A signal on CD41H cells and CD41 cells (upper graph). Intensity of VEGF-A signal was quantified with ImageJ software from photomicrographs stained with anti-VEGF-A (green channel) and anti-CD41 (red channel), as shown in Supporting Fig. 2. Dots represent the mean intensity value obtained for each cell. Mean intensity and SEM are shown as the thin horizontal and vertical lines (n = 61 and 57 for CD41H cells and CD41 cells, respectively). P values calculated with the unpaired t test are shown; **P < 0.01. Bottom graph represents the percentage of CD41H cells and CD41 cells that display bright VEGF-A signals (black, VEGF-A intensity ≥100), weak VEGF-A signals (hatched, VEGF-A intensity between 25 and 99), or are negative for VEGF-A (white, VEGF-A intensity ≤24). Data were obtained from four images corresponding to two independent experiments. P values correspond to the chi-square test; *P < 0.05. (C) Quantification of ALB expression by quantitative real-time PCR performed on samples from purified E11.5 FL CD49fHCD41H MKPs and CD49fD cells as well as from total FL cells at E11.5, E12.5, and E15.5, and on adult liver (Li). Values were normalized to those of the GαS gene, and those obtained for E11.5 FL cells were used as a reference. Bars represent the mean ± SEM (n = 4): *P < 0.05; **P < 0.01. (D) TTR, AFP, ALB, AAT, platelet factor-4 (PF4), and GαS detection by PCR in CD45CD41H cells from YS, PBL, and AGM and FL at E11.5. CD45CD41H cells from FL were purified by double-sorting procedures. Water and E11.5 FL cells were used as negative (C) and positive (C+) controls, respectively.

We next verified by IF whether CD41H MKPs from FL expressed the hepatocyte nuclear factors (HNFs), HNF-1, HNF-3β, and HNF-4α, which are essential for the expression of most hepatocyte genes. In preparations from unpurified E11.5 FL cells, and from purified c-KitDCD45 and CD49fHCD41H cells, there was only a weak punctuate nuclear HNF-4α and HNF-1 signal in CD41H cells (Fig. 5A and Supporting Fig. 5), and no staining for HNF-3β was observed (not shown). By contrast, brighter homogeneous signals were detected in the nuclei of CD49fDCD41 cells. In addition, no surface expression of hepatic glucose transporter type 2 (GLUT2) was detected in CD49fHCD41H MKPs (Fig. 5B). Therefore, the ALB protein detected in CD49fHCD41H MKPs from the E11.5 FL is most probably accumulated by endocytosis.

Figure 5.

HNF-4α, GLUT2, and Dlk expression by CD49fH cells present in the FL of E11.5 mouse embryos. (A) Dual IF with anti-CD41 to trace MKP (red channel) and anti-HNF-4α to identify ALB-producing cells (green channel). Nuclei were counterstained with TOPRO-3; samples were analyzed on a Leica DMRD confocal microscope (Leica, Wetzlar, Germany), and images were processed with ImageJ software. Photomicrographs show individual channels for each Ab. Images correspond to cytospin preparations of FL cells at E11.5 (FL11.5). (B and C) Representative dot plots of the flow cytometry of FL cells at E11.5 performed with the Abs indicated. Fluorescence scales are logarithmic, and the FSC scale is linear. Cells negative for Ter119 and CD45 were electronically gated and subsequently analyzed. Gates inside the dot plots show the anti-CD49f and anti-c-Kit staining that identifies the CD49fH and CD49fD subpopulations analyzed with anti-GLUT2, anti-CD41/Dlk, or anti-CD13/Dlk in the histograms and dot plots indicated by the arrows. (B) Staining with the anti-GLUT2 Ab is shown in the histograms. ISO signal is shown overlayed (dotted line). Numbers represent percentage of cells (mean ± SEM; n = 3). (C) Rectangles in the dot plots showing anti-CD41/Dlk staining define the Dlk+ populations. Numbers represent percentage of cells (mean ± SEM; n = 10). (D) Bar graph shows the quantification of Dlk+ cells per liver at E10.5 (white bars) and E11.5 (black bars) among the CD49fH and CD49fD subpopulations. Data are expressed as the mean ± SEM, where n = 3 and 10 for the FL samples at E10.5 and E11.5, respectively. ***P < 0.001.

To further clarify the relationship between FL MKPs and HeP, the Dlk/CD13 markers used to define liver stem/progenitor cells17 were analyzed on electronically gated CD49fHCD41H and CD49fD cells from FL (Fig. 5C,D). We found that CD49fD cells contained most Dlk+CD13+ cells (1,291 ± 389 cells/FL), whereas CD49fHCD41H MKPs contained only 62.5 ± 9.8 cells/FL (n = 10) of Dlk+ cells. Taken together, these results reinforce the idea that FL CD49fHCD41H MKPs are distinct to HeP, even though they share some characteristics of hepatoepithelial and endothelial cells.

CD49fHCD41H MKPs From E11.5 FL Drive the In Vitro Development of Hepatoepithelial Layers in Culture.

The c-KitDCD45 population contained HeP that can establish hepatoepithelial layers in vitro.10 Because the subpopulation of CD49fH CD41H cells present in the c-KitDCD45 HeP appear to belong to the MK lineage, and the remaining CD49fD cells express hepatoepithelial transcripts and contain Dlk+CD13+ cells, we reasoned that these CD49fD cells may represent the true HeP present in the FL at E11.5. To investigate this hypothesis, we cultured purified c-KitDCD45CD49fD (CD49fD) cells after removing c-KitDCD45CD49fH (CD49fH) cells by FACS. In the absence of the CD49fH population, CD49fD cells could not grow in culture on any of the substrates tested (uncoated, collagen I, laminin, or fibronectin), and after 3 days in culture, most of them adopted a small, round appearance of apoptotic cells (Supporting Fig. 6). When CD49fH cells were seeded along with CD49fD cells, the mix of the purified subpopulations formed hepatoepithelial layers, as did cultures of total purified c-KitDCD45 cells (Fig. 6A). These cultured cells expressed HNF-4α (Supporting Fig. 6). We concluded that the presence of CD49fHCD41H MKPs was required for CD49fD HeP to grow in vitro. To determine whether this process was mediated by direct cell-to-cell contacts or by soluble factors, we cultured the purified CD49fH and CD49fD populations in transwells (Fig. 6B). Again, epithelial layers developed when both subpopulations were grown together in the upper chamber of transwell plates. By contrast, the growth of CD49fD HeP seeded in the upper chamber was poorer when the CD49fH MKPs were plated in the bottom chamber, despite the normal in vitro development of proplatelets in these cultures. The addition of conditioned medium from CD49fHCD41H cells to CD49fD cultures promoted a limited growth of hepatoepithelial layers (Supporting Fig. 6), in agreement with the fact that supernatants from complete FL cell cultures or growth-promoting factors need to be added to adult or FL-derived liver progenitors to generate hepatoepithelial layers in vitro.11-13, 18

Figure 6.

Interaction between CD49fH MKPs and CD49fD HeP in the FL at E11.5. Purified c-KitDCD45Ter119 cells, c-KitDCD45Ter119CD49fH MKPs (CD49fH) and c-KitDCD45Ter119CD49fD HeP (CD49fD) were obtained by FACS from E11.5 FL cell suspensions. Purified cells were seeded on Col I-coated plates. (A) Bright-field images from 7-day cultures of CD49fD HeP cocultured with CD49fH MKPs (left) and from cultures of the complete c-KitDCD45 population (right). (B) Representative images from one experiment of three performed on purified cells cultured for 7 days in chambers of transwell plates and seeded as indicated below the images. Framed populations are cells shown in each image. Dashed lines highlight the epithelial layers that develop (C and D) ALB, and KDR expression was analyzed by quantitative real-time PCR (as in Fig. 4C). Values from CD49fD HeP cultures were used as a reference (value = 1). Bars represent the mean ± SEM (n = 3): *P < 0.05; ***P < 0.001. (C) After 7 days in culture, cells grown in upper transwell chambers were harvested and their ALB expression was analyzed (left bar graph). H, CD49fH cells; D, CD49fD cells. ALB expression in cells grown on Col I-coated plates for 7 days in the presence of 50% conditioned medium (CM) obtained from cultures of CD49fH MKPs, serotonin (5-HT), or VEGF-A (right bar graph). (D) KDR expression on cells grown on collagen I–coated plates in the presence or absence of VEGF-A (left bar graph). ALB expression on cells grown on Col I-coated plates in the presence of anti-mouse VEGF-A or ISO (right bar graph). Results are shown as the expression relative to values obtained for cultures of CD49fD cells in the absence of Abs (100%).

The induction of ALB and AAT expression was considered evidence of hepatocyte differentiation in our cultures (Fig. 6C and Supporting Fig. 6). The greatest increase in ALB expression was induced when both CD49fH MKP and CD49fD HeP cells were grown together in the same chamber (4.9-fold). Conversely, when these two populations were separated by a membrane, ALB expression increased similar to that induced in CD49fD cultures to which conditioned medium was added (2.1- and 2.5-fold, respectively). Serotonin and VEGF were both detected in MKs and platelets and may play a role in hepatocyte growth and regeneration after liver injury.19, 20 Indeed, FL CD41H cells express the highest levels of VEGF-A in the FL (Fig. 4B). It has also been reported that maternal serotonin promotes embryonic FL growth.21 Although serotonin neither induced hepatoepithelial layer formation nor increased ALB expression in our system, VEGF-A induced both effects to a similar extent to that observed after the addition of conditioned medium, as well as inducing an increase in VEGFR2/KDR expression (Fig. 6D). By contrast, the addition of anti-VEGF Abs to c-KitDCD45 cells reduced ALB levels in cells of these cultures. Thus, in addition to the cell-to-cell contacts required for complete development of hepatoepithelial layers, our data indicate that soluble factors derived from MK and, in particular, VEGF-A are involved in the growth of ALB-producing cells.

Finally, the involvement of MKPs in establishing the hepatoblast niche in vivo was suggested by the close localization of both MKPs (as CD41H) and HeP (as ALB++) in vivo at E11.5, as demonstrated by the contact observed between MKPs and ALB++ cells (Fig. 7A,C) and between MKPs and the more-abundant c-Kit+ subpopulation or other MKPs (Fig. 7B,C). These data show that direct cellular contacts between MKs and HeP occur physiologically, and strongly suggest that MKs may facilitate the development of the hepatoepithelial liver compartment.

Figure 7.

Contacts between CD49fHCD41H MKPs and CD49fD HeP are observed in ex vivo samples from the E11.5 FL. Dual IF in FL tissue sections at E11.5 using the Abs indicated and counterstaining nuclei with TOPRO-3 (blue). Photomicrographs show merged channels for each staining. Analyses were performed with a Zeiss LSM 510 confocal microscope (Carl Zeiss Microscopy GmbH, Jena, Germany) and were processed with ImageJ software. (A) Photomicrographs show images obtained with Abs against CD41 (red) and ALB (green). Contacts between CD41H cells and ALB++ cells are shown in yellow (white arrows). (B) Microphotographs of images obtained with anti-c-Kit (red) and anti-CD41 (green) Abs. (C) Relative contacts between MKPs and HeP (ALB, black), c-Kit+ cells (hatched), and other MKPs (CD41, white). Radius of cells from each population, their frequency, and the number of contacts between them were counted by two independent observers (mean ± SEM; a total of 17 × 103 cells were analyzed in 40 photomicrographs). These data were used to calculate the absolute numbers of each population, their total surface area, and the normalized contacts/cell surface area.


During FL morphogenesis in the postgastrulation embryo, a liver-specific progenitor (the hepatoblast) can be identified by its capacity to differentiate to both hepatocytes and cholangiocytes.10, 11 The phenotype of the early HeP at E11.5 has been defined as c-KitD/−CD45Ter119, with variable levels of CD49f expression, together with other markers, such as the hepatocyte growth factor (HGF) receptor (c-Met) and Dlk.10-12, 18 However, postnatal liver progenitors have been described as CD49fH.13 Our results demonstrate that at E11.5, the c-KitDCD45Ter119 liver progenitors contain two subsets of cells: one defined as CD49fDCD41 that could represent HePs and another phenotype defined as CD49fHCD41H, suggesting an MK lineage. These CD49fHCD41H FL cells respond to ADP and thrombin stimulation, rapidly differentiating in vitro, as described for embryonic MEPs in semisolid assays.5, 6 Indeed, after 24 hours in culture, cytoplasmic elongations develop and proplatelets are generated in a process dependent upon the reorganization of the actin cytoskeleton, similar to that described in mature MKs.8, 22 Our observation that proplatelets are present in vivo under physiological conditions in isolated cellular FL preparations is particularly relevant until, as recently, proplatelet development by MKs was demonstrated to occur in vivo after TPO treatment.23

The CD49fHCD41H MKPs found in the FL represent a potential source of pure MKPs that are readily isolated by FACS or immunomagnetic methods, in contrast to the BM clonogenic MK population that is relatively small.4, 16 As observed in adult BM MKs, c-KitDCD49fHCD41H cells from E11.5 FL express several integrin receptors. The engagement of α4β1, and not αVβ3, has been proposed to enhance TPO-induced megakaryopoiesis.24 In FL c-KitDCD49fH CD41H cells, we observed weaker α4 expression in relation to the αV chain, which may be related to the TPO-independent maturation of these cells in vitro. This observation is consistent with findings from c-Mpl-deficient mice, in which MK generation occurs in a TPO-independent manner before E9.5 in the YS and before E14.5 in the FL.6, 25 The α6 integrin chain (CD49f) associates with either β1 (CD29) or β4 (CD104) integrin chains to form receptors for laminin and kalininis, respectively, and has been implicated in adhesion and in vivo homing. This chain is expressed by HSCs and myeloid progenitors in E14.5 FL and BM,26 among other cells, and by MKs and platelets generated in vitro.27 However, to the best of our knowledge, CD49f has not commonly been associated with MKs ex vivo. In contrast to the adult BM MK-lineage cells (including mature MKs) and other embryonic myeloid cells, embryonic CD49fHCD41H MKPs do not express the hematopoietic marker (CD45).

Several of the cell-surface markers presented by the CD49fHCD41H MKPs present in the FL at E11.5 differ from those in other hematopoietic niches (manuscript in preparation). Specifically, these cells express endothelial and hepatoepithelial proteins, the latter probably being taken up by endocytosis in the liver (where they would be produced by the HeP), because CD49fHCD41H MKPs appear to only weakly or not at all express HNF-1, HNF-4α, and HNF-3β. Similarly, most CD49fHCD41H MKPs are DlkCD13, indicating that they are not liver stem/progenitor cells. The possible relationship between the few CD49fHDlk+ cells and the more-abundant CD49fD Dlk+CD13+ cells will require further analysis. These embryonic CD49fHCD41H MKPs preferentially contact ALB+ hepatoblasts in vivo and may contribute to specific hepatocyte-developmental niches, as proposed for adult MKs in the establishment of plasma cell niches in the BM.16 In vitro, CD49fHCD41H MKPs stimulate the development of CD49fD HeP. Moreover, in transwell cultures, hepatospecific genes are up-regulated in immature CD49fD HeP in response to direct cell contact and CD49fHCD41H cell-derived soluble factors, in particular VEGF-A, which is produced most strongly by CD49fHCD41H cells. In fact, although VEGFR2/KDR is weakly expressed at E11.5 ex vivo by CD49fD HeP, its expression is up-regulated in vitro after the addition of VEGF-A. MKs produce VEGF,28 which participates in the endothelial organization of the vasculature, vasculogenesis, and blood island formation, and fulfils other nonvascular roles in the morphogenesis of adult organs and stem cell niches.16, 29-31 In addition to their role in hemostasis, platelets, the end product of MK differentiation, are involved in liver regeneration and hepatocyte proliferation through direct contact as well as the release of HGF, insulin growth factor, and VEGF.32, 33 Indeed, they are also involved in several other biological processes, including the spread of hematogenic tumor cells,34 vessel remodeling in the newborn,35 and the separation of blood and lymphatic circulation during development.36, 37 In conclusion, the data presented here describe the precise phenotypic identification of embryonic CD49fHCD41H MKPs. Our findings propose interesting new tools to study the role of MKs in tissue regeneration and strongly support a role new for CD49fHCD41H MKPs in the development of the FL, involving the action both of cellular contacts and VEGF-A.


The authors thank Beatriz Palacios, Fernando Martínez, and Carmen Prado for their technical assistance and help with the animal care and Mark Sefton for his editorial assistance.