The visceral endoderm comprises a single layered cup-shaped epithelium, which encases the epiblast and extraembryonic ectoderm of the postimplantation embryo and will later form the endodermal layer of the yolk sac (reviewed by Rossant,1986). The visceral endoderm functions in nutrient uptake and transport and also expresses a large number of secreted serum proteins and transcription factors in common with the adult gut and liver. Consequently, it is thought to nurture the embryo, both before and after placental connections are established. More recent studies suggest that the visceral endoderm also plays an active role in the morphogenesis and patterning of the early embryo (Belaoussoff,1998a; Dyer et al.,2001; Lu et al.,2001). The definitive endoderm comprises a population of multipotent progenitor cells representing one of the primary germ layers generated during gastrulation (Tam et al.,2001; Hogan and Zaret,2002). The definitive endoderm gives rise to the major cell types of the digestive tract and associated organs, including the liver and pancreas. Little is known of the genetic determinants that regulate the specification, differentiation, and morphogenesis of the definitive endoderm in mice.
Recent advances in the isolation of proteins that naturally fluoresce and the refinement of techniques for in vivo microscopy, offer unprecedented opportunities for studying cellular and molecular events within living, intact embryos (Hadjantonakis et al.,2003; Passamaneck et al.,2006). The generation of strains of mice expressing genetically encoded fluorescent proteins in defined lineages now permits live imaging of tissue morphogenesis in a wild-type and mutant context. To develop a mouse model that will permit real-time imaging and isolation of primitive and definitive endodermal cells, we have used a 7.6-kb promoter/enhancer from the mouse Alpha-fetoprotein (Afp) gene to drive expression of fluorescent reporters. Afp is the most abundant serum protein of the mammalian embryo (Tilghman,1985; Spear,1999). An oncodevelopmental protein that is expressed at high levels in the embryonic yolk sac and fetal liver, its expression decreases dramatically after birth (Andrews et al.,1982; Tilghman and Belayew,1982). Whereas only trace amounts of Afp are synthesized in the adult liver, owing to a dominant repression domain (Emerson et al.,1992; Ramesh et al.,1995) ensconced within a much larger 5′ upstream regulatory region (Krumlauf et al.,1985; Hammer et al.,1987), Afp gene expression is reactivated during conditions involving rapid hepatocyte proliferation, such as liver regeneration or tumorigenesis. Afp has provided an excellent model system for studying the tissue- and developmental stage-specific regulation of gene expression (reviewed by Spear,1999). In particular, the 7.6-kb Afp promoter/enhancer fragment has been characterized previously in the yolk sac and fetal liver of transgenic mice (Krumlauf et al.,1985; Hammer et al.,1987) and during the directed differentiation of transgenic embryonic stem (ES) cells (Paparella et al.,2002).
In this study, we have generated lines of Tg(Afp-GFP) mice (hereafter termed Afp-GFP) that express green fluorescent protein (GFP) specifically within the visceral endoderm, its derivative the yolk sac endoderm, hepatocytes of the fetal liver, and the epithelium of the fetal gut and pancreas. We demonstrate the utility of these lines for tagging and tracking of endodermal cells of the primitive and definitive lineages of the mouse embryo, using microscopic imaging and flow cytometry.
RESULTS AND DISCUSSION
GFP Expression in Explanted Blastocysts
A single line of Afp-KGFP and two lines of Afp-EGFP transgenic mice (see cartoon in Fig. 1A) were analyzed. No differences in the spatial or temporal localization of GFP expression during embryogenesis were observed among the Afp-KGFP and Afp-EGFP lines. Therefore, the data have been pooled and the lines referred to generically as AFp-GFP. To determine the onset of transgene expression as visualized by green fluorescence, we recovered embryos at the blastocyst stage at embryonic day (E) 3.5 to E4.5, coincident with the specification of primitive endoderm (reviewed by Rossant,1986). We were unable to visualize GFP in any of the preimplantation stages recovered (E0.5–E4.5) in either hemizygous or homozygous Afp-GFP transgenic embryos (Fig. 1B). However, if E4.5 blastocysts were plated onto tissue culture dishes to promote attachment and outgrowth, GFP expression was initiated within 24 hr (Fig. 1C–E). The onset of fluorescence was documented in individual blastocyst outgrowths using time-lapse imaging (Fig. 1C–E). In a total of 42 embryos examined from 8 litters, 19 blastocysts were GFP-positive. Of interest, GFP-positive cells were never found interspersed within outgrowths. They were usually observed as single, and occasionally paired, cohorts of cells located internally in the outgrowth. Because blastocyst attachment to a dish usually occurs within the first 24 hr after plating, it was likely that the GFP first became visible a few hours after attachment. Furthermore, high-resolution time-lapse imaging of individual blastocysts revealed that a single cell activated expression of the transgene and then divided to form a clone of fluorescing cells, as opposed to a model wherein onset of expression occurs in multiple neighboring cells (data not shown). The morphology of GFP-expressing cells, when examined at high magnification, resembled that of primitive endoderm stem cells (XEN cells, Kunath et al.,2005). In particular, we observed the dynamic, unidirectional pseudopodial-like projections characteristic of XEN cells (Fig. 1F–M). Because XEN cells have been reported to express markers of both parietal and visceral endoderm, we would expect them to express Afp.
GFP Expression in Postimplantation Embryos
By early postimplantation stages, GFP expression was readily detectable in both hemizygous and homozygous transgenic Afp-GFP embryos (Fig. 2). Confocal imaging of whole embryos confirmed that fluorescence was localized to the visceral endoderm (Fig. 2B,C,F,G). At E5.5, while the distal visceral endoderm overlying the epiblast expressed GFP, the proximal visceral endoderm overlying the extraembryonic ectoderm was devoid of fluorescence (Fig. 2A–D,I). By E6.5, coincident with the onset of gastrulation, expression of GFP had spread more proximally, into the extraembryonic visceral endoderm (Fig. 2E–H,J). This observation is in general agreement with a previous report that signals emanating from the extraembryonic ectoderm repress Afp expression in the overlying visceral endoderm (Dziadek,1978). Repression is relieved later, when mesoderm generated during gastrulation has infiltrated the extraembryonic region (Dziadek,1978; Dziadek and Adamson,1978).
In gastrula stage embryos, the domain of GFP expression corresponded with the localization of Afp transcripts as it came to occupy the entire visceral endoderm and subsequently the endoderm component of the vitelline yolk sac (Fig. 3). Coincident with formation of definitive endoderm during gastrulation, the domains of expression of GFP and Afp RNA were displaced proximally (Fig. 3A–C). High-resolution imaging and three-dimensional (3D) reconstruction of yolk sac tissue counterstained with a vital nuclear dye (DRAQ5) confirmed the localization of green fluorescence to cells of the yolk sac endoderm (Fig. 3J–L).
By midgestation (E9.5), fluorescence was detected throughout the yolk sac endoderm (Fig. 4A) and within the liver primordium (Fig. 4B–D). Both sites of transgene expression were in concordance with the localization of Afp transcripts in stage-matched embryos (Fig. 4E–H). By E11.5, fluorescence was maintained within the yolk sac endoderm and liver (Fig. 4I,J). However, sectioning revealed heterogenous fluorescence in a dispersed population of hepatocytes (Fig. 4K). High-magnification imaging revealed that these GFP-expressing cells exhibited a hepatoblastic morphology, with between one and four protrusions per cell (Fig. 4L). Time-lapse imaging of cultured transverse sections of liver from E11.5 embryos established the dynamic nature of these protrusions (Fig. 4M–P). In contrast, Afp transcripts were identified in all cells of the liver and in yolk sac endoderm (Fig. 4Q–T). At E13.5, the Afp-GFP transgene was widely expressed in the liver (Fig. 5A).
Endoderm Specificity of GFP Expression
Endoderm (hepatocyte) specificity was confirmed by immunostaining of dispersed cells from E14.5 liver (Fig. 5B). All GFP(+) cells expressed endogenous Afp (Fig. 5B) and hepatocyte nuclear transcription factor (HNF4, Duncan et al.,1994; Fig. 5C). However, only a subset (∼10%) of Afp-expressing cells were GFP(+), with over 680 Afp-expressing cells examined (Fig. 5B). The expression of the Afp-GFP transgene within a subpopulation of liver cells may reflect heterogeneity in the onset and level of expression. In addition, the constructs generated for this study lacked an intronic enhancer and alternative promoter known to contribute to Afp expression in the yolk sac and fetal liver (Scohy et al.,2000; Gabant et al.,2002).
Consistent with the hematopoietic activity of the fetal liver, immunostaining of cryosections revealed the presence of Ter119+ erythroid (middle panel) and F4/80(+) macrophage (right panel) populations (Fig. 5C). GFP was expressed in a distinct population of cells with characteristic hepatoblastic morphology (Fig. 5C, right panel). The large size of these cells is evident from the contour plot of GFP fluorescence vs. forward scatter (Fig. 5D).
Cell Surface Phenotype of GFP-Expressing Cells in the Embryo
In addition to their utility for imaging the developing embryo, the GFP reporter mouse lines also allow the isolation and analysis of cells in which the transgene is expressed, using fluorescence-activated cell sorting (FACS). To determine whether GFP could be detected by FACS, single-cell suspensions of Afp-GFP transgenic livers were prepared from E14.5 embryos and analyzed by flow cytometry. Approximately 10% of the fetal liver cells expressed GFP at high levels (Fig. 5D).
To examine the surface phenotype of the GFP(+) liver cells, single-cell suspensions were stained with antibodies against a variety of surface antigens and analyzed by FACS. Fetal liver cells have been demonstrated to express several adhesion molecules, including α4 integrin, α5 integrin, and CD44 (Kawakami et al.,1999); all three were expressed on a majority of GFP(+) cells (Fig. 5E). CD24 and CD147/Basigin function in integrin signaling (Hahne et al.,1994; Berditchevski et al.,1997; Sammar et al.,1997; Cho et al.,2001) and were also expressed on the GFP(+) population (Fig. 5E). Heterogeneity within the GFP(+) population was suggested by the presence of VCAM-1, a receptor for the α4β1 integrin complex VLA-4 (Hemler et al.,1987) on half of the GFP(+) cells (Fig. 5E). Fetal hepatic progenitors have been reported to express α6 integrin (Suzuki et al.,2002) and low levels of c-kit (Minguet et al.,2003). A subset of GFP(+) cells in the fetal liver expressed c-kit (8%, Fig. 5E) but not α6 integrin (not shown) and may represent a somewhat more mature population of cells. The C regulatory protein CD55 protects the fetus from lysis by maternal complement (Simpson et al.,1993) and was expressed at high levels in most of the GFP(+) cells of the liver (Fig. 5E). Expression of hematopoietic and endothelial markers (CD41, CD45, Ter119, F4/80, CD34, Sca-1 CD31, MECA-32, and Flk1) was not detected within the GFP(+) population (not shown).
GFP Expression in Late Stage Yolk Sac and Definitive Endoderm Derivatives
In late stage embryos, strong expression of GFP was observed in the yolk sac (Fig. 6B,C), liver, and pancreas (Fig. 6D,E). In the yolk sac, Ter119 staining identified erythroid cells in blood vessels adjacent to the GFP(+) visceral endoderm (Fig. 6G). The yolk sac (Fig. 6H) and fetal liver endoderm (Fig. 6D) and the definitive endodermal layer of the intestine (Fig. 6I) and pancreas (Fig. 6J) displayed overlapping expression of GFP and the epithelial marker E-cadherin. Ectopic expression of GFP (but not endogenous Afp mRNA) was detected in the brain in all lines (Fig. 6F and data not shown).
Afp-GFP Transgenic Mice: A Novel Tool for Analysis of Primitive and Definitive Endoderm.
In summary, the Afp-GFP transgenic mice permit noninvasive imaging of the visceral endoderm, the endoderm of the yolk sac, and several definitive endoderm derivatives. They provide a novel tool for studying the dynamic behavior of visceral endoderm cells, including their displacement and/or migration during gastrulation (Srinivas et al.,2004). The bright endodermal fluorescence will facilitate the isolation of primitive and definitive endodermal cell populations and their derivatives by microdissection and their identification, quantitation, and purification by flow cytometry of living tissue. In view of the known reactivation of Afp, an oncofetal tumor marker, during liver tumorigenesis, these lines should allow a readout in real time in preclinical models of human hepatocellular carcinoma and may represent a useful model for evaluating antitumor drugs for the treatment of liver cancer.
A limited number of transgenic mouse lines that express fluorescent reporters in endodermal lineages are currently available. The Hex-GFP line expresses GFP within a subpopulation of visceral endoderm, the prospective anterior visceral endoderm (AVE), and has been used to track the migration of these cells (Srinivas et al.,2004). However, Hex is also expressed in a subset of definitive endoderm, endothelial, primitive erythroid, and endocardial cells, all mesodermal derivatives (Thomas et al.,1998). Otx2 upstream regulatory sequences have been shown to drive regionalized GFP expression in the visceral endoderm (distally, at E5.5, and in the AVE, by E6.25; Kimura et al.,2000). In Pdx1-GFP transgenic mice, GFP fluorescence was observed in the developing pancreas, stomach, and duodenum (Gu et al.,2004). Pdx1 is not expressed in the visceral endoderm and to date GFP fluorescence was not described in this tissue (Gu et al.,2004). To our knowledge, the Afp-GFP transgenic mice reported here are currently the most specific lines available for the identification, visualization, and characterization of both primitive and definitive endodermal lineages in the developing mouse embryo.
Generation of Afp-GFP Transgenic Mice
The Afp-KGFP transgene construct was built in a pSP73 (Promega, Madison, WI) backbone and contained an EcoRI/SalI fragment containing a 7.6-kb enhancer from the α-fetoprotein (Afp) gene (Hammer et al.,1987; Spear and Tilghman,1990); jellyfish green fluorescent protein (KGFP) coding sequences amplified using the polymerase chain reaction (PCR) to include HindIII and XbaI sites at the 5′ and 3′ termini, respectively; and sequences from the human ϵ-globin gene (+474 to +1746) containing part of exon 2, IVS 2, exon 3, 3′ untranslated sequences, and a polyadenylation signal. The Afp enhancer was excised from pUC9-Afp-Dd (Spear and Tilghman,1990). The ϵ-globin sequences were excised as a XbaI/EcoRI fragment from −179ϵlacZμLCR (Belaoussoff,1998b) and inserted by blunt ligation into the PvuII site of pSP73. For microinjection into embryos, plasmid DNAs were digested with EcoRI and XhoI restriction enzymes and the eukaryotic portions purified using standard methods (Nagy et al.,2003). Pronuclear injection into C57BL/6 × C3H (B6C3) F1 hybrid embryos was performed by the Mount Sinai Mouse Genetics Shared Resource Facility. The mice were maintained as hemizygotes on an ICR outbred background. Transgenic males were crossed with ICR females for all studies. The morning of detection of the vaginal plug was taken as day 0.5 of gestation.
To generate the Afp-EGFP vector, the 7.6-kb Afp enhancer was cloned into plasmid phsp68-EGFP-SV40pA (A.-K.H., unpublished results), containing the hsp68 minimal promoter, the yellow-shifted EGFP variant fluorescent protein (Clontech, Inc., Palo Alto, CA) coding sequence, and an SV40 polyadenylation sequence. The eukaryotic insert DNA was excised with NotI. After purification using routine protocols (Nagy et al.,2003), DNA was injected into C57BL/6 zygotes at the Memorial Sloan-Kettering Cancer Center Transgenic Core Facility.
Preimplantation embryos were recovered in M2 medium and cultured in KSOM Embryo Culture Medium (Chemicon Specialty Media, Temecula, CA) in an organ culture dish (BD Falcon, catalog no. 353037) or, for live imaging, under mineral oil in a MatTek glass bottom dish (catalog no. P35G-1.5-14C), at 37°C, 5% CO2. Postimplantation embryos and organs were dissected in N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES) buffered Dulbecco's Modified Eagle's Medium (DMEM) containing 10% fetal bovine serum (FBS). During live imaging, specimens were cultured in 50% rat serum, 50% DMEM/F12 supplemented with L-glutamine (2 mM). For fixation, embryos were dissected in phosphate-buffered saline (PBS) and subsequently transferred into 4% paraformaldehyde in PBS (PFA/PBS). All data presented are from hemizygous single transgenic animals. Homozygous animals displayed more robust fluorescence (not shown).
Vibrating Microtome Sectioning and Counterstaining
Specimens were either embedded and sectioned immediately after dissection or were fixed in 4% PFA/PBS for 2 to 12 hr, washed in PBS, and embedded in 4% low-melt agarose, 5% sucrose in PBS. Blocks were cut out of embedding molds, trimmed using a razor blade, and then mounted onto a vibrating microtome chuck (Leica VT1000S) using superglue. Sections were cut at a thickness of 200 μm. Embryos were counterstained with 7.5 mM DRAQ5 (Alexis Biochemicals, San Diego, CA) in 50% DMEM:50% rat serum.
In Situ Hybridization and Immunochemistry
In situ hybridization of biotinylated RNA probes to whole-mount embryos and vibrating microtome sections was performed according to standard protocols (Nagy et al.,2003). Embryos were dissected in PBS and fixed in 4% PFA/PBS for 12 to 24 hr, then washed in PBS before being processed for in situ hybridization as either whole-mounts or 200-μm-thick sections. Hybridizations were performed using an Afp probe (Law and Dugaiczyk,1981).
For immunocytochemistry, fetal livers from E14.5 Afp-GFP embryos were dissociated by simple trituration and filtration through a 70-μm cell strainer (BD Falcon, catalog no. 352350) to produce single cell suspensions. Cells were centrifuged onto Fisherbrand Superfrost/Plus glass slides for 3 min at 600 rpm (Cytospin3 cytocentrifuge, Shandon, Inc., Pittsburgh, PA) and allowed to air dry overnight. Cytospins were fixed for 10 min at room temperature (RT) in 4% PFA/PBS, blocked for 1 hr at RT in PBS containing 0.1% Carnation instant nonfat dry milk and 0.05% Tween-20 (Sigma) (PBSMT). Slides were then stained overnight at 4°C with anti-Afp antibody (Afp AB-2, rabbit polyclonal, NeoMarkers, Lab Vision, Fremont, CA, catalog no. RB-365-A1 or Afp (C-19), goat polyclonal IgG, Santa Cruz Biotechnology, catalog no. sc-8108) diluted 1:400 in PBSMT. The two anti-Afp sera produced comparable patterns of staining, but background fluorescence was lower for the NeoMarkers antibody, which was used for the experiments shown in the figures shown here. Anti-Hnf4α was from Santa Cruz Biotechnology (catalog no. sc-6556). The following day, slides were washed three times with PBS containing 0.05% Tween-20 (PBST) and incubated with anti-rabbit IgG-Alexa568 (Molecular Probes, Eugene, OR) diluted 1:400 in PBST for 1 hr at RT. The cytospins were washed three times in PBS and mounted using Vectashield Mounting Medium with DAPI (catalog no. H-1200; Vector Labs, Burlingame, CA).
For immunohistochemistry, embryos and individual organs were harvested and processed as described previously (Fraser et al.,2005). They were then transferred to Tissue-Tek OCT embedding medium (Sakura Finetek, Torrance, CA) for 3 hr, placed into embedding moulds, and snap-frozen in isopentane chilled in a liquid nitrogen bath. Cryosections (10 μm) were cut using a Leica CM3050 cryostat, allowed to adhere to Superfrost/Plus glass slides, and air-dried overnight. They were then blocked for 1 hr at RT in PBSMT and incubated for 1 hr at RT with primary antibody. Primary antibodies used were Ter119 (BD Biosciences, San Diego, CA), F4/80 (EBiosciences, CA), and E-cadherin (Zymed, South San Francisco, CA). Slides were then washed three times in PBST and incubated for 1 hr at RT with anti-rat IgG-Alexa568 secondary antibody (Molecular Probes) diluted 1:400 in PBST. Slides were then washed in PBS and mounted as described above.
Laser scanning confocal data were acquired using a Zeiss LSM510 META laser scanning confocal mounted on a Zeiss Axiovert 200M microscope. Whole embryos were kept in PBS in MatTek glass bottom dishes during imaging. Fluorescence was excited with a 488-nm Argon laser (GFP) or a 633-nm HeNe laser (DRAQ5). Objectives used were as follows: C-apochromat 40×/NA1.2, plan-apochromat 20×/NA0.75, fluar 5×/NA0.25, or plan-apochromat 63×/NA1.4. Optical section thicknesses ranged from 0.2 to 2 μM. Widefield images were acquired using a Axiocam MRC (Zeiss) camera mounted onto a Leica MZ16FA stereodissecting microscope. For live imaging and time-lapse experiments, embryos were maintained at 37°C in humidified chamber, under 5% CO2 (Solent Sci., UK).
Single cell preparations of E14.5 fetal livers were prepared as described above and incubated with fluorescently conjugated antibodies to detect surface antigen expression. Conjugated antibodies were purchased from EBiosciences (EBio) or BDBiosciences (BD). Phycoerythrin-conjugated antibodies were as follows: α6 integrin (BD), Ter119 (EBio), CD31 (EBio), CD34 (BD), CD41 (BD), CD55 (EBio), CD147 (EBio), and Sca-1 (BD). CD44 and c-kit antibodies (BD) were directly conjugated to allophycocyanin (APC). Biotin-conjugated antibodies were α4- and α5-integrin, from BD; and CD45, F4/80, MECA-32, VCAM-1, and Flk1, from EBio. Biotin was detected by using streptavidin–APC (BD). Cells were washed in FACS buffer (10% FBS in PBS) and resuspended in FACS buffer containing 0.01% propidium iodide to allow exclusion of dead cells. Cells were analyzed using a FACSCalibur four-color analyzer (Becton Dickinson) in the Mount Sinai Flow Cytometry Shared Research Facility.
We thank Chantal Lackan for assistance during the course of this work, and Drs Kevin Kelley and Liz Lacy for helpful discussions. The Afp-Dd construct was a generous gift from Dr. Shirley Tilghman.