|Home | About | Journals | Submit | Contact Us | Français|
Primordial germ cells (PGCs) are the progenitors of reproductive cells in metazoans and are an important model for the study of cell migration in vivo. Previous reports have suggested that Hedgehog (Hh) protein acts as a chemoattractant for PGC migration in the Drosophila embryo and that downstream signaling proteins such as Patched (Ptc) and Smoothened (Smo) are required for PGC localization to somatic gonadal precursors. Here we interrogate whether Hh signaling is required for PGC migration in vertebrates, using the zebrafish as a model system. We find that cyclopamine, an inhibitor of Hh signaling, causes strong defects in the migration of PGCs in the zebrafish embryo. However, these defects are not due to inhibition of Smoothened (Smo) by cyclopamine; rather, we find that neither maternal nor zygotic Smo is required for PGC migration in the zebrafish embryo. Cyclopamine instead acts independently of Smo to decrease the motility of zebrafish PGCs, in part by dysregulating cell adhesion and uncoupling cell polarization and translocation. These results demonstrate that Hh signaling is not required for zebrafish PGC migration, and underscore the importance of regulated cell-cell adhesion for cell migration in vivo.
In metazoans, germline tissue is required for the transmission of genetic information from one generation to the next. Consistent with the importance of this function, many behaviors of early germline tissue are conserved across species. Primordial germ cells (PGCs), progenitors of the germline lineage, are typically specified in locations distal to the future gonad field and therefore must migrate extensively through somatic tissues during embryogenesis. Once PGCs arrive at the presumptive gonad site, they associate with somatic precursors of the reproductive organs and ultimately differentiate into sperm or ova.
The molecular mechanisms that regulate germ cell specification, migration, and differentiation are not completely understood, but studies of Drosophila and zebrafish have provided key insights into invertebrate and vertebrate PGC development, respectively (Kunwar et al., 2006; Raz, 2003). In fruitflies, PGCs are initially specified by maternally derived mRNAs and other germ plasm components that are localized to the posterior end of the early embryo (Lehmann and Nusslein-Volhard, 1986). During gastrulation, the Drosophila PGCs are carried with associated somatic cells to the posterior midgut, after which they traverse the midgut epithelium and attach to the overlying mesoderm (Warrior, 1994). The PGCs then migrate bilaterally toward clusters of somatic gonadal precursor cells, with which they coalesce to form the gonad (Boyle et al., 1997). Zebrafish PGCs are similarly specified by maternal germ plasm determinants, in this case localized to the cleavage planes at the two- and four-cell stages (Weidinger et al., 1999). By the 1000-cell stage (3 hours post fertilization (hpf)), the zebrafish embryo contains four PGCs in a square-like configuration that is randomly oriented with respect to the dorsal-ventral axis. The PGCs then begin to proliferate, and between 4 and 5 hpf they transition from round, immotile cells into a polarized migratory population (Blaser et al., 2005), localizing to the anterior and lateral boundaries of the mesoderm during gastrulation (Weidinger et al., 1999). The PGCs continue to migrate toward an intermediate region bordering the mesoderm during somitogenesis, until 25 to 50 cells are bilaterally restricted to the anterior end of the yolk extension (between the eighth and tenth somites) at 24 hpf.
The long-range migration undertaken by PGCs has garnered particular interest not only as an integral step in gonad development, but also as model system for the study of directed cell movements in vivo. Previous studies have revealed several factors that are necessary for both PGC migration and survival. For example, germ plasm in Drosophila and zebrafish contains a conserved suite of PGC-specific mRNAs, including vasa (Hay et al., 1988; Yoon et al., 1997) and nanos (Koprunner et al., 2001; Wang and Lehmann, 1991), and embryos lacking these transcripts or the proteins they encode exhibit mislocalized PGCs that are ultimately lost. In fruitflies, PGC motility and maintenance is further regulated by lipid phosphate phosphatases encoded by the wunen genes, which appear to produce a repellent signal (Renault et al., 2004). Zebrafish PGC development is also regulated by the dead end (dnd) gene, which is essential for PGC migration and survival (Weidinger et al., 2003).
A second class of factors is required for PGC migration but not survival, regulating either the basic motility of these progenitor cells or their chemotaxis. For instance, zebrafish PGCs deficient for phosphatidylinositol 3-kinase (PI3K) signaling have reduced speed but maintain their capacity for directed migration (Dumstrei et al., 2004). Similarly, E-cadherin is downregulated in zebrafish PGCs as they begin to migrate, and its aberrant upregulation reduces PGC migration speed but not directionality (Blaser et al., 2005). Other genes are required for the chemotaxis of PGCs but not their general motility. Such factors include the zebrafish chemoattractant stromal-derived factor 1a (sdf1a) (Doitsidou et al., 2002) and its receptor cxcr4b (Knaut et al., 2003). PGC chemotaxis in both Drosophila and zebrafish also requires the enzyme 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMG-CoAR). In fruitflies, HMG-CoAR is selectively expressed in somatic gonadal precursor cells during the second phase of PGC migration, and loss of this gene in columbus mutants results in PGC mislocalization (Van Doren et al., 1998). Conversely, cells ectopically expressing HMG-CoAR can attract migratory PGCs. Since this enzyme is the rate-limiting step in the mevalonate pathway, which is specifically utilized in fruitflies to synthesize isoprenoids, it has been hypothesized that an isoprenylated protein is required for PGC attraction (Santos and Lehmann, 2004). Consistent with this model, mislocalized PGCs are observed in Drosophila lacking farnesyl-diphosphate synthase, geranylgeranyl diphosphate synthase, or geranylgeranyl transferase 1, three enzymes that act downstream of HMG-CoAR in the biosynthesis of geranylgeranylated proteins, and ectopic expression of either synthase gene is sufficient to induce fruitfly PGC chemotaxis (Santos and Lehmann, 2004). A conserved regulatory mechanism may also operate in zebrafish, as embryos treated with pharmacological inhibitors of either HMG-CoAR or geranylgeranyl transferase 1 exhibit germ cell migraton defects (Thorpe et al., 2004).
Other studies have implicated the Hedgehog (Hh) pathway as a regulator of PGC migration in fruitflies. Drosophila Hh signaling is initiated by the secreted Hh ligand, which acts through its receptor Patched (Ptc) and the transmembrane protein Smoothened (Smo) to activate Hh target gene expression by the Cubitus interruptus (Ci) transcription factor (Ingham and McMahon, 2001). Hh pathway activity is also promoted by the serine-threonine kinase Fused (Fu) and repressed by protein kinase A (PKA). It was first reported that Drosophila embryos lacking maternal ptc, smo, pka, or fu exhibit PGC migration defects, suggesting that inappropriate Hh pathway upregulation or downregulation disrupts germ cell migration to the somatic gonadal precursor cells (Deshpande et al., 2001). In addition to PGC mislocalization, loss of either maternal ptc or pka caused premature PGC aggregation, perhaps due to precocious acquisition of a developmental signal that is normally produced by the presumptive gonad. Since hh reporter activity was also observed in somatic gonadal precursor cells and ectopic hh expression could partially re-direct PGC migration, it was hypothesized that the Hh protein acts as a PGC chemoattractant, possibly through the activation of downstream signaling components (Deshpande et al., 2001).
More recently it has been proposed that HMG-CoAR regulates Drosophila PGC migration by potentiating the release of Hh protein from Somatic gonadal precursor cells (Deshpande and Schedl, 2005), perhaps by promoting geranylation of the G protein submit Gγ1 (Deshpande et al., 2009). Embryos lacking zygotic HMG-CoAR or Gγ1 activity exhibit significantly reduced Hh protein movement, and HMG-CoAR overexpression in Hh-producing cells causes an expansion of the Hh ligand distribution. These perturbations also induced changes in Smo localization and expression of the Hh target gene wingless that are consistent with HMG-CoAR- and Gγ1-dependent release of active Hh ligand. Consistent with the Hh chemoattractant model, it has further been observed that fruitflies lacking tout-velu (ttv), a critical regulator of heparan sulfate proteoglycan biosynthesis and Hh protein movement, exhibit PGC migration defects (Deshpande et al., 2007).
Whether Hh ligands serve a similar chemoattractant role in other organisms is not known. Cellular and molecular similarities between germ cell ontogeny in Drosophila and zebrafish suggest that the latter organism would be an ideal model system for investigating a possible role for Hh signaling in vertebrate PGC migration. We therefore sought to determine whether Hh signaling is required for PGC migration in zebrafish, using both pharmacological and genetic perturbations of this developmental pathway. Our studies exploit the amenability of zebrafish embryos to genetic manipulations, chemical perturbations, and optical imaging, which provides an opportunity to analyze vertebrate PGC migration in real-time and with subcellular resolution (Raz, 2003). Here we report that cyclopamine, an inhibitor of Hh signaling that directly targets Smo, elicits a strong defect in PGC motility. This chemical perturbation involves both germline and somatic cell lineages, but does not affect PGC chemotaxis or maturation. Cyclopamine, however, exerts these effects through Smo-independent mechanisms, including dysregulation of PGC cell adhesion. In contrast to previous observations in Drosophila, zebrafish embryos lacking both maternal and zygotic Smo exhibit normal PGC migration, indicating that Hh proteins do not act as a PGC chemoattractant in this system. Our studies also raise the possibility that other organismal and cellular phenotypes associated with cyclopamine treatment may involve Hh pathway-independent mechanisms of action.
Zebrafish (Danio rerio) of the wildtype AB genetic background and strains heterozygous for mutant alleles of smo (smuhi1640 and smub577), sonic hedgehog (syutbx392), gli1 (dtrts269), gli2a (yotty119), or dispatched1 (contm15a) were obtained from either the Zebrafish International Resource Center, the Tübingen Zebrafish Stock Center, or individual laboratories. Tg(kop:EGFP-F-nanos1-3’UTR) zebrafish, which selectively express farnesylated, enhanced GFP in PGCs, have been previously described (Blaser et al., 2006). All zebrafish strains were maintained, raised, and staged according to standard protocols (Kimmel et al., 1995; Westerfield, 1995). Embryos were obtained by natural matings and cultured at 28 °C unless otherwise noted.
Cyclopamine treatments were initiated at the one-cell stage with a 100 μM solution in E3 medium (Nusslein-Volhard and Dahm, 2002) unless otherwise noted. All cyclopamine-containing media were made from a stock solution in ethanol, such that the final ethanol concentration was 0.2% (v/v). An equivalent amount of ethanol in E3 medium was used a vehicle control. Time course experiments were performed by culturing embryos in ethanol-containing E3 medium until the indicated time points, after which they were transferred into medium containing 100 μM cyclopamine. Cyclopamine washout experiments were conducted by incubating the embryos in cyclopamine-containing E3 medium and washing them extensively (at least three times) with ethanol-containing E3 medium at end of each treatment regimen. The embryos were then cultured further in ethanol-containing E3 medium.
One- and two-color whole-mount in situ hybridization protocols were performed as described previously (Hauptmann and Gerster, 1994; Jowett and Lettice, 1994). The following digoxigenin-labeled probes were used: vasa (Olsen et al., 1997; Yoon et al., 1997), nanos1 (Koprunner et al., 2001), sdf1a (Doitsidou et al., 2002), and ziwi (Houwing et al., 2007). The vasa probe was used to label PGCs, which were scored according to their location within or outside of the presumptive gonad site in the lateral plate mesoderm. At least 400 PGCs (approximately 15 embryos) were typically scored to determine the percentage of mislocalized cells.
Whole-mount immunostaining of zebrafish embryos was performed according to standard protocols (Macdonald, 1999; Nusslein-Volhard and Dahm, 2002) using the following primary antibodies: rabbit polyclonal anti-Vasa (Knaut et al., 2000) (1:5,000 dilution), rabbit polyclonal anti-green fluorescent protein (GFP) (Invitrogen/Molecular Probes, 1:500 dilution), mouse monoclonal anti-E-cadherin (clone 36, BD Biosciences; 1:500 dilution), mouse monoclonal anti-β-actin antibody (clone AC-15, Sigma; 1:1,000 dilution), and mouse monoclonal anti-α-tubulin (clone DM1a, Sigma; 1:500 dilution). Alexa Fluor 488- and Alexa Fluor 594-conjugated secondary antibodies (Invitrogen/Molecular Probes) were used at a 1:1,000 dilution.
Capped sense RNA for the following constructs was synthesized with the MessageMachine kit (Ambion) and microinjected into one-cell stage embryos: sdf1a-globin-3’UTR (Doitsidou et al., 2002), EGFP-nanos1-3’UTR (Koprunner et al., 2001), and vasa-dsRedex-nanos1-3’UTR (Strasser et al., 2008). The nanos1-3’UTR sequence was used to direct protein expression to the PGCs.
Antisense morpholinos (Gene-Tools) targeting the following genes were microinjected into one-cell stage embryos: sdf1a (5-CTACTACGATCACTTTGAGATCCAT-3’) (Doitsidou et al., 2002), sdf1b (5’-TTGCTATCCATGCCAAGAGCGAGTG-3’) (Boldajipour et al., 2008), E-cadherin (5’-AAAGTCTTACCTGAAAAAGAAAAAC-3’) (Shimizu et al., 2005), E-cadherin five-base mismatch (5’-AAACTCATACGTGAAAAACAAATAC-3’), and smo (5’-CGCTTGGAGGACATCTTGGAGACGC-3’) (Robu et al., 2007).
Zebrafish Smo cDNA was subcloned into the pCS105 vector through ClaI and NheI sites using the PCR primers 5’-TATATAAGCTTATCGATCCACCATGTCCTCCAAGCGCCCT-3’ and 5’-AAGCTAGCGAAAAATCTGAGTCAGCATCCAATAGCT-3’. Smo−/− murine embryonic fibroblasts (MEFs) (Ma et al., 2002) were cultured in DMEM containing 10% fetal bovine serume (FBS) to approximately 50% confluency and then co-transfected with the zebrafish Smo expression construct, a Gli-dependent firefly luciferase reporter (Sasaki et al., 1997), and a constitutive Renilla luciferase reporter (pRLSV40, Promega) using Fugene 6 transfection reagent (Roche) according to the manufacturer’s protocols. After two days, the cells were cultured in DMEM containing 0.5% FBS, the N-terminal domain of Sonic Hedgehog protein (Shh-N) (Chen et al., 2002b), and various concentrations of cyclopamine for an additional 30 h. Firefly and Renilla luciferase activities in the Smo−/− MEFs were then determined using a dual luciferase kit (Promega) according to the manufacturer’s protocols.
Donor embryos were prepared by injecting them at the one-cell stage with Alexa Fluor 594-dextran (1.8 ng/embryo; 10,000 MW, Invitrogen/Molecular Probes) and EGFP-nanos1-3’UTR mRNA (200 pg/embryo) (Koprunner et al., 2001). Donor and host embryos were then cultured in E3 medium containing either 100 μM cyclopamine or an ethanol vehicle control as appropriate. Donor and host embryos were raised at 28 °C and 24 °C, respectively, so that they would reach the proper stages for transplantation at the same time. Just prior to cell harvesting and transplantation, the embryos were dechorionated with pronase and then washed extensively with E2 medium (Nusslein-Volhard and Dahm, 2002).
Transplantations were performed in E2 medium using an Eppendorf CellTram Vario. Needles were made from 1.0-mm glass capillaries pulled on a Sutter P-87 pipette puller, and their tips were broken manually with forceps. During transplantation, PGCs were visualized by EGFP fluorescence with a Leica MZFLIII stereoscope, and one to three PGCs were removed from shield-stage donor embryos with as little surrounding tissue as possible (usually 10 to 25 cells). This tissue was then implanted into the margin of sphere-stage host embryos. The host embryos were cultured further in cyclopamine-free E3 medium to 24 hpf, and PGCs were visualized by EGFP fluorescence. Transplanted PGCs were scored according to their location within or outside of the presumptive gonad, defined to be lateral plate mesoderm between the eighth and tenth somites.
PGC directed migration was assessed as previously described (Blaser et al., 2005). Donor embryos from wildtype AB adult zebrafish were microinjected at the one-cell stage with sdf1a MO (0.4 pmol/embryo), sdf1b MO (0.4 pmol/embryo), Alexa Fluor 568-dextran (150 pg/embryo; 10,000 MW, Invitrogen/Molecular Probes), and synthetic MO-resistant sdf1a-globin- 3’UTR mRNA (240 pg/embryo). Tg(kop:EGFP-F-nanos1-3’UTR) host embryos were microinjected at the one-cell stage with the sdf1a and sdf1b MOs and then treated with 100 μM cyclopamine or an ethanol control until the transplantation procedure.
Somatic cells were harvested from donor embryos at 30% epiboly (5 hpf) and transplanted into host embryos at 50% epiboly (5.5 hpf). The host embryos were then maintained in cyclopamine-containing medium and imaged by time-lapse fluorescence microscopy (6 to 7 hpf).
Time-lapse imaging of PGC migration in Tg(kop:EGFP-F-nanos1-3’UTR) embryos was performed using an upright Leica DM4500B compound microscope equipped with a narrow-pass GFP filterset, a 10X objective, and a deep-cooled, charge-coupled device camera (Retiga SRV) controlled by ImagePro software (MediaCybernetics). Tg(kop:EGFP-F-nanos1-3’UTR) embryos were immobilized in an agarose template during late gastrulation and early segmentation (8 to 12 hpf), and PGCs in the dorsolateral region were imaged at a rate of one frame/min for at least 120 min with manual refocusing as necessary. Cyclopamine-treated embryos were cultured in drug-containing E3 medium throughout each time-lapse experiment. Embryos cultured in E3 medium containing the ethanol vehicle alone were imaged as controls.
PGCs from the image stacks were tracked manually in ImagePro, with three tracks generated for each cell to minimize user error. In addition, morphological features in the unlabeled somatic tissue were tracked to account for background morphogenetic movements in the gastrulating embryo. Approximating the zebrafish embryo as a two-dimensional plane, tracks were then exported as coordinate data. Somatic tissue coordinate data were subtracted from the PGC coordinate data to correct for background movements, and PGC migration speeds were averaged over three consecutive frames. To quantify the directed migration of individual PGCs, the uncorrected PGC coordinate data were used to calculate the change in angle trajectory between frames. Angle changes were then averaged among the three tracks over three consecutive frames to provide a mean angle change for the PGC at the center frame.
Individual PGCs were assigned a morphology state of either “elongated” or “round,” and a phase state of “run” or “tumble” for each frame (Reichman-Fried et al., 2004). The run phase for these studies was defined to be a motile state with an elongated cellular morphology. Adhesive interactions between germ cells were also recorded, using the plasma membrane-localized green fluorescence in Tg(kop:EGFP-F-nanos1-3’UTR) PGCs to identify regions of cell-cell contact.
Maternal-zygotic mutant zebrafish were generated by germline replacement as previously described (Ciruna et al., 2002). To create MZsmuhi1640 zebrafish, progeny from intercrosses of smuhi1640 heterozygotes were injected with EGFP-nanos1-3’UTR mRNA (200 pg/embryo) and then used at the sphere stage as donor for transplantation into sphere-stage wildtype hosts. About 100 to 200 cells were transplanted into each host, and successful transplants were scored at 24 hpf by the presence of EGFP fluorescence at the presumptive gonad site (typically 15-25% of the transplanted hosts). The chimera were then raised to adulthood and mated to smuhi1640 heterozygotes to ascertain gamete genotype. One female bearing smuhi1640 oocytes and five males bearing smuhi1640 sperm were isolated from 69 chimeric adults.
To create MZsmub577 zebrafish, donor embryos were generated by intercrosses of smub577 heterozygotes, and wildtype embryos were used as hosts (Thomas et al., 2008). The donor embryos were injected with EGFP-nanos1-3’UTR mRNA, and at 24 hpf the host embryos were screened for the presence of donor-derived, EGFP-expressing PGCs. The genotype of the corresponding donor embryo for each host was determined phenotypically at this stage. Three females bearing smub577 oocytes and six males bearing smub577 sperm were successfully isolated from 67 chimeric adults.
Both MZsmuhi1640 and MZsmub577 embryos exhibit morphological phenotypes similar to those observed in zygotic smu mutants, including U-shaped somites, ventral body curvature, and circulation defects.
To conduct western blot analyses of E-cadherin expression, zebrafish embryos (6 or 24 hpf) were manually dechorionated and dissociated in ice-cold buffer (5 mM HEPES, pH 7.0, 100 mM NaCl, 5 mM KCl, 1% (w/v) PEG-20,000) containing a protease inhibitor mixture (Roche) and 1 mM phenylmethanesulfonylfluoride. The dissociated cells were then collected by centrifugation (300 g, 4 °C) and disrupted in sodium dodecyl sulfate-polyacrylamide gel electrophoresis loading buffer with sonication. The resulting cellular lysates were then separated by gel electrophoresis and transferred to polyvinylidene fluoride membranes. At least 20 embryos were used for each experimental condition, with lysate obtained from approximately five embryos analyzed in each gel electrophoresis lane. E-cadherin immunoblotting was conducted by chemiluminescence detection (Pierce, SuperSignal West Pico), using either a mouse monoclonal anti-E-cadherin antibody (clone 36, BD Biosciences; 1:5000 dilution) or a rabbit polyclonal anti-E-cadherin (1:10,000 dilution) (Babb and Marrs, 2004). Western blot analyses of actin levels were performed as loading control, using a mouse monoclonal anti-β-actin antibody (clone AC-15, Sigma; 1:10000 dilution).
Since previous studies have implicated the Hh ligand and its downstream effectors as regulators of Drosophila PGC migration (Deshpande and Schedl, 2005; Deshpande et al., 2007; Deshpande et al., 2001), we sought to investigate whether Hh pathway components are required for PGC migration in vertebrates. We focused on zebrafish PGCs as a model system, since the genetic tractability, pharmacological sensitivity, and optical transparency of zebrafish embryos have made them ideally suited for analyzing germline development (Raz, 2003). In addition, a number of Hh signaling proteins in zebrafish have been cloned, and their expression patterns during embryogenesis have been analyzed. For example, the essential Hh signal transducer smo is provided maternally, uniformly transcribed during gastrulation, and then preferentially expressed in anterior tissues by 24 hpf (Chen et al., 2001; Varga et al., 2001).
We first treated zebrafish embryos with cyclopamine, a small-molecule inhibitor of Hh signaling that directly inhibits Smo (Chen et al., 2002a). We found that continuous exposure to 100 μM cyclopamine starting at the one-cell stage caused PGC mislocalization throughout the embryos by 24 hpf, as determined by the distribution of vasa-positive cells (Fig. 1, A-B). Treatment of zebrafish embryos with simpler organic amines such as piperidine and diethylamine at identical concentrations did not yield this PGC phenotype (data not shown), confirming that the cyclopamine-dependent PGC mislocalization is dependent upon its specific three-dimensional structure and not due to it basicity. The degree of PGC mislocalization at the highest doses of cyclopamine is at least comparable to that observed with the statins, pharmacological inhibitors of HMG-CoAR that perturb PGC migration in zebrafish (Thorpe et al., 2004).
We next analyzed the dose-dependence of cyclopamine action on PGC localization. We observed that this teratogen has a median effective concentration (EC50) of approximately 30 μM, with the percentage of mislocalized PGCs reaching a plateau by 100 μM concentrations (Fig. 1C). This activity profile is similar to that reported for cyclopamine-dependent Hh pathway inhibition in zebrafish as measured by ptc1 transcript levels (Wolff et al., 2003). Since cyclopamine has a median inhibitory concentration (IC50) of 300 nM toward Hh pathway activation in murine cell lines (Taipale et al., 2000), this antagonist could be either less potent against zebrafish Smo than mouse Smo or less able to penetrate zebrafish embryos than cultured cells. To distinguish between these two possibilities, we assessed the ability of cyclopamine to inhibit zebrafish Smo in cultured cells. Embryonic fibroblasts derived from Smo−/− mice were transiently transfected with zebrafish smo, which restores the ability of these cells to activate Hh target gene expression in response to Shh-N-conditioned medium. As determined by a co-transfected Gli-dependent firefly luciferase reporter, cyclopamine was very potent at inhibiting zebrafish Smo-mediated Shh signaling, with an IC50 of approximately 15 nM (Fig. 1D). Since cyclopamine should exhibit a similar activity against zebrafish Smo in vivo, we conclude that the embryonic concentration of cyclopamine is also nanomolar when drug concentrations in the culture medium are in the micromolar range. This 1,000-fold difference in cyclopamine concentration is likely due to poor tissue penetrance or rapid efflux from the embryo.
In principle, cyclopamine could disrupt the targeting of zebrafish PGCs to the presumptive gonad at any point during the first 24 h of germ cell development. Specified by maternally provided determinants, zebrafish PGCs are initially round, immotile cells. Between 4 and 5 hpf, they exhibit a polarized morphology and extend protrusions in a stochastic manner, until they begin to migrate actively toward regions of sdf1a expression in the developing embryo. This migratory phase continues during gastrulation and somitogenesis, with the PGCs reaching the presumptive gonad by 24 hpf (Blaser et al., 2005; Weidinger et al., 1999). To define the temporal window during which cyclopamine affects PGC localization, we varied the developmental stages at which we initiated exposure to this teratogen. We observed that cyclopamine exerted its effects on early gastrulation, as gauged by the resulting distribution of vasa-positive cells at 24 hpf (Fig. 1E). Treatment with cyclopamine had its strongest effects when initiated by 2 hpf, and compound administration after 6 hpf had little effect on the localization of these progenitor cells, even though the bulk of PGC migration to the presumptive gonad site occurs during this time.
We also performed complementary experiments in which cyclopamine was added to one-cell stage embryos and then washed out a different time points. The effects of these treatments on PGC localization at 24 hpf was assayed, and cyclopamine treatment from 0 to 6 hpf was required for the strongest degree of PGC mislocalization. Teratogen exposure from 0 to 3 hpf only had a small effect on PGC targeting to the presumptive gonad. Taken together, these data indicate that the critical time window for cyclopamine-induced PGC mislocalization is between 3 and 6 hpf, a time during which PGCs are only beginning their migration toward sdf1a-expressing somatic tissues.
Since cyclopamine acts early during PGC development to perturb cell migration, we investigated whether its effects were cell-autonomous with respect to these progenitor cells. We performed cell transplantation studies in which we transferred PGCs with as few somatic cells as possible from shield-stage (6 hpf) donor embryos to sphere-stage (4 hpf) hosts (Fig. 2A). This transplantation regimen allowed us to analyze PGC function after the critical time window for cyclopamine action has passed (3 to 6 hpf).
The donor embryos were injected with Alexa Fluor 594-dextran to provide a red fluorescent marker for all donor cells and EGFP-nanos1-3’UTR mRNA to selectively label the PGCs with green fluorescence. After transplantation, the donor-derived cells are widely distributed such that the PGCs are almost exclusively in contact with host tissue (Supplementary Movie 1). Using EGFP fluorescence we then scored the locations of donor PGCs in the host embryo at 24 hpf. When the donor and host embryos were both treated with an ethanol vehicle before transplantation, we observed a basal level of PGC mislocalization in response to the transplantation regimen alone (Fig. 2, B and B’). As expected, the extent of PGC mislocalization significantly increased when both donor and host embryos were treated with cyclopamine prior to transplantation (Fig. 2, C and C’).
Using this system, we could then determine whether cyclopamine exerts its effects on the PGCs, somatic tissues, or both, by administering cyclopamine to only the donor or host embryos. When only the donor embryo was treated with cyclopamine, we observed a moderate increase in PGC mislocalization over basal levels (Fig. 2, D and D’). However, when the host embryo was treated with cyclopamine before transplantation, a greater fraction of PGCs failed to reach the presumptive gonad site (Fig. 2, E and E’). Quantitative analyses of these data confirm that cyclopamine acts primarily through the soma to perturb PGC migration (Fig. 2F), although the teratogen directly impairs the progenitor cells to some extent. This mechanism of action does not appear to involve sdf1a, which acts as a PGC chemoattractant (Doitsidou et al., 2002), since cyclopamine does not significantly alter sdf1a expression during gastrulation or somitogenesis (Fig. 2, G-L).
To further investigate the mechanism by which cyclopamine causes PGC mislocalization, we used time-lapse imaging to study PGC migration in Tg(kop:EGFP-F-nanos1-3’UTR) embryos, which harbor a transgene that selectively expresses farnsylated EGFP in these cells (Blaser et al., 2005). We observed the migration of PGCs during late gastrulation and early segmentation (8 to 12 hpf), since cyclopamine-treated PGCs first digress from sdf1a-expressing tissues at these stages (Supplementary Fig. 1). In embryos treated with an ethanol vehicle control, the PGCs migrated independently (Supplementary Movie 2), following long trajectories that often circled back upon themselves but traveled significant distances from the origin (Fig. 3A). In contrast, embryos continuously treated with cyclopamine starting at the one-cell stage had PGCs that exhibited slower motility and adhered to each other frequently (Supplementary Movie 3). The cyclopamine-treated PGCs also followed short trajectories that did not extend far from the origin (Fig. 3A).
Zebrafish PGCs undergo a series of alternating phases in their migration, which have been denoted “running” and “tumbling” in analogy to bacterial movement (Reichman-Fried et al., 2004). Run phases are characterized by cell elongation and polarization, as well as a burst of motion along the vector of cell polarity. In contrast, tumble phases intersperse run phases and are characterized by rapidly changing polarization vectors with minimal overall migration. Each phase persists for a few minutes.
To study the migratory behavior of cyclopamine-treated PGCs more closely, we analyzed PGC speed, directionality, and morphology in embryos exposed to the teratogen or an ethanol vehicle control (Fig. 3B). Continuous treatment with cyclopamine beginning at the one-cell stage not only diminished PGC speed (Fig. 3C), but also reduced the fraction of time spent in run phases (Fig. 3D). Interestingly, cyclopamine did not change the fraction of time PGCs exhibit an elongated, polarized morphology (Fig. 3D), which is nearly always associated with the run phase in wildtype embryos (see Fig. 3B). These seemingly contradictory observations can be reconciled if cyclopamine decouples cell polarization and translocation, and consistent with this model, the cyclopamine-treated PGCs were often immotile even when they were elongated and polarized. The average duration of these abnormal polarized phases was greater than that of polarized phases in ethanol-treated embryos (Fig. 3E), but the frequency with which cyclopamine-treated embryos switched between non-polarized and polarized morphologies was reduced in comparison to control cells (Fig. 3F and Supplementary Movies 4 and 5). Collectively, these migration defects resulted in an overall decrease in cell motility.
In addition to characterizing the PGC motility defect, we used real-time imaging of their migration to verify the temporal window of action of this teratogen. When embryos were exposed to cyclopamine from 0 to 6 hpf, their PGCs exhibited movement defects during late gastrulation that were similar to those of PGCs treated with this steroid alkaloid throughout embryogenesis (data not shown). However, PGCs migrated normally in embryos treated with cyclopamine from 6 hpf onward (data not shown). These combined results confirm that cyclopamine action prior to 6 hpf is required for its effects on PGC migration.
To determine whether PGC chemotaxis is inhibited by cyclopamine in addition to motility, we observed the ability of cyclopamine-treated PGCs to respond to transplanted tissue overexpressing the chemoattractant sdf1a (Doitsidou et al., 2002). PGCs treated with either cyclopamine or an ethanol vehicle control were able to migrate short distances to the transplanted source of Sdf1a protein (Fig. 3G and Supplementary Movies 6 and 7). Thus, the PGC mislocalization observed in cyclopamine-treated embryos appears to result from defects in general cell motility rather than a loss of directed migration.
Since several genes required for PGC migration are also necessary for the maintenance of these progenitor cells, we investigated whether cyclopamine causes any discernible defects in PGC morphology or maturation. The prototypical PGC marker genes vasa and nanos1 are inherited from maternal stores and expressed in PGCs from the earliest stages of their specification (Koprunner et al., 2001; Olsen et al., 1997; Yoon et al., 1997). Cyclopamine does not have any effect on the transcript levels of vasa (see Fig. 1, A-B) or nanos1 (Fig. 4, A-B), indicating that cyclopamine-treated PGCs retain their germ plasm-derived components.
Vasa protein expression provides a later indicator of PGC development, as this gene is translated soon after PGC fate commitment is established (4 hpf) and is localized to discrete perinuclear granules (Knaut et al., 2000). Using an antibody raised against zebrafish Vasa (Knaut et al., 2000), we failed to detect any change in the expression of endogenous Vasa protein upon cyclopamine treatment (Fig. 4, C-D). The cellular organization of Vasa-containing perinuclear granules was also not perturbed by this teratogen, as determined by selectively expressing fluorescently tagged Vasa (Vasa-dsRedex) in PGCs (Fig. 4, E-F). Nor did cyclopamine appear to perturb PGC maturation, which begins after these cells reach the presumptive gonad site (24 hpf). One of the earliest known markers of germ cell maturation is the expression of ziwi, the zebrafish ortholog of the piwi family of RNA-interacting proteins (Houwing et al., 2007). Even though PGCs are displaced from the presumptive gonad in cyclopamine-treated embryos, ziwi transcripts can be detected in these cells by 48 hpf, the same time frame as wildtype PGCs (Fig. 4, G-H). In addition, cyclopamine treatment did not elicit any detectable change in the number of PGCs throughout their development, indicating that it does not impede PGC proliferation or survival (Supplementary Fig. 2). These observations indicate that this teratogen does not disrupt the expression of germ-cell specific genes in PGCs, the subcellular organization of these cells, or their maturation. Rather, the effects of cyclopmine appear to be specific to PGC motility.
Although cyclopamine is widely used as a specific inhibitor of the Hh pathway, it is possible that this teratogen inhibits PGC migration in a Hh pathway-independent manner. We did not observe PGC defects in mutant embryos lacking zygotic function of Shh (syutbx392) (Schauerte et al., 1998), Gli1 (dtrts269) (Karlstrom et al., 2003), Gli2a (yotty119) (Karlstrom et al., 1999), Dispatched1 (contm15a) (Nakano et al., 2004), or Smo itself (smuhi1640) (Chen et al., 2001) (Supplementary Fig. 3). However, the failure to detect PGC defects in these zygotically mutant embryos could be due to sufficient levels of maternally supplied factors during the temporal window of cyclopamine action on PGCs (3 to 6 hpf). For example, maternal smo transcripts are abundant during early embryogenesis (Chen et al., 2001; Varga et al., 2001). To address this possibility, we also assessed PGC localization in zebrafish embryos microinjected with a MO that targets the translational start site in smo transcripts. PGC successfully homed to the presumptive gonad site in these embryos, although the morphant phenotypes were generally less severe than those of smuhi1640 mutants (Supplementary Fig. 3). The smo MO may therefore may be limited by incomplete efficacy or slow kinetics of action. It is also possible that early stage embryos contain maternally derived Smo protein, which would be unaffected by these antisense oligonucleotides.
To unequivocally determine whether Smo is required for zebrafish PGC migration, we depleted both maternal and zygotic smo in embryos using the germline transplantation technique (Ciruna et al., 2002). We used progeny resulting from smuhi1640/+ intercrosses as germline donors, since this allele is the most complete loss of Smo function of the existing smu alleles (Chen et al., 2001). From our transplanted embryos, we isolated one adult chimeric female zebrafish that carried smuhi1640 ova and four chimeric males that carried smuhi1640 sperm. Crossing the female to any of the four males resulted in 100% of the progeny exhibiting U-shaped somites, ventral body curvature, circulation defects, and other smu mutant phenotypes, (Fig. 5A), thus confirming that these adult zebrafish contain germline exclusively derived from smuhi1640 progenitor cells.
Progency resulting from crossing the chimeric adults lack both maternal and zygotic smo (MZsmuhi1640), providing the appropriate genetic background for assessing a requirement for Smo in zebrafish PGC development. We did not observe PGC mislocalization in MZsmuhi1640 embryos; moreover, PGC migration defects could be induced in these mutants upon exposure to cyclopamine (Fig. 5, B-C). These results were further confirmed using independently generated chimeric adults containing ova or sperm homozygous for the smub577 allele (Varga et al., 2001) (Supplementary Fig. 4). Thus, the ability of cyclopamine to perturb PGC migration is not due to Smo inhibition, and Smo-dependent processes such as Hh signaling are dispensible for proper PGC migration in zebrafish.
Since cyclopamine does not perturb PGC migration by inhibiting Smo, we sought to determine the mechanism of action of this teratogen. The specific perturbation of PGC speed by cyclopamine rather than chemotaxis or maturation directed us to analyze functional components of the cell motility apparatus. We first investigated whether cyclopamine perturbs actin or tubulin polymers in PGCs, as these cytoskeletal structures have been shown to be necessary for zebrafish PGC migration (Blaser et al., 2006). Global changes in actin (data not shown) or tubulin (Fig. 4, I-J) cytoskeletal architecture were not observed in cyclopamine-treated PGCs at 6 hpf, during which these cells are normally undergoing extensive migration. These results indicate that cyclopamine does not grossly disrupt the PGC cytoskeleton.
Other regulators of zebrafish PGC migration include cell adhesion molecules such as E-cadherin, in analogy to other migratory cell populations (Nakagawa and Takeichi, 1998; Niewiadomska et al., 1999; Nishimura et al., 1999). E-cadherin is downregulated during the onset of PGC motility, presumably to achieve cell adhesive properties with the surrounding somatic tissues that are appropriate for migration (Blaser et al., 2005). We therefore assessed whether cyclopamine inhibits PGC migration by altering cell adhesive interactions in the zebrafish embryo. Our time-lapse movies of PGC migration revealed that cell-cell contacts between PGCs persisted for longer durations in embryos exposed to cyclopamine than those treated with an ethanol vehicle control alone (Fig. 6, A-B). In principle, this increased contact time could either cause or reflect the decreased speed of cyclopamine-treated PGCs. In support of the former option, we observed that reduction of E-cadherin expression by MO knockdown partially rescued the cyclopamine-induced PGC defect when teratogen and E-cadherin MO doses were appropriately matched. For example, we achieved a partial rescue of PGC migration in embryos treated with 50 μM cyclopamine upon the microjection of E-cadherin MO at a dose of 300 pg/embryo (Fig. 6, C-E and Supplementary Fig. 5). PGC migration in embryos microinjected with the E-cadherin MO at doses greater than 500 pg/embryo could not be analyzed since these conditions induced gastrulation defects analogous to those observed in half-baked mutants, which lack zygotic E-cadherin (Kane et al., 2005) (data not shown). In contrast, a five-base mismatch control MO did not reduce E-cadherin expression levels and did not mitigate cyclopamine-induced PGC mislocalization (Supplementary Fig. 5).
During the course of our time-lapse experiments, we also observed that PGCs in cyclopamine-treated embryos occasionally fragment during their migration, perhaps reflecting an uncoupling of translocation and adhesion disassembly during cell migration (Supplementary Movie 8). Collectively, these observations suggest that cell adhesive forces are dysregulated upon cyclopamine exposure. We did not observe, however, cyclopamine-dependent changes in E-cadherin expression levels or subcellular localization, as determined by western blot analysis and whole-mount immunofluoresence (Fig. 6, F-H). Thus, while the abnormal cell adhesion is likely to be at least one contributing mechanism by which cyclopamine perturbs PGC migration, this teratogen appears to affect cell adhesion molecules other than E-cadherin. Alternatively, cyclopamine might modify the intrinsic activity of E-cadherin proteins, perhaps increasing their adhesive properties.
The molecular mechanisms that regulate PGC specification, migration, and differentiation are remarkably conserved across species. For example, nanos and vasa play critical roles in PGC specification in Drosophila and zebrafish (Hay et al., 1988; Koprunner et al., 2001; Wang and Lehmann, 1991; Yoon et al., 1997). Mammalian orthologs of nanos and vasa are also required for germ cell development (Fujiwara et al., 1994; Tsuda et al., 2003), although these progenitor cells are specified by inductive signals such as the bone morphogenetic proteins (Lawson et al., 1999; Ying et al., 2000; Ying and Zhao, 2001). Orthologs of the dnd gene promote PGC migration and survival in zebrafish and mice (Weidinger et al., 2003; Youngren et al., 2005), and HMG-CoAR is required for PGC chemotaxis in fruitfies and zebrafish (Thorpe et al., 2004; Van Doren et al., 1998). In addition, members of the piwi family of RNA-interactors are critical regulators of germline maintenance in fruitflies, zebrafish, and mice (Cox et al., 1998; Deng and Lin, 2002; Houwing et al., 2007; Kuramochi-Miyagawa et al., 2004).
Previous studies have implicated Hh signaling proteins in the regulation of Drosphila PGC migration (Deshpande and Schedl, 2005; Deshpande et al., 2007; Deshpande et al., 2001). It has been reported that fruitfly embryos lacking maternal expression of the Hh pathway regulators ptc, smo, fu, or pka exhibit mislocalized PGCs and that germ cells can be recruited to sites of ectopic hh expression (Deshpande et al., 2001). More recent findings have suggested that HMG-CoAR and its downstream target Gγ1 promote Hh protein secretion and movement (Deshpande et al., 2009; Deshpande and Schedl, 2005), thereby supporting a model in which somatic gonadal precursor cells produce Hh protein to attract migratory PGCs to the presumptive gonad site. Given the molecular similarities of PGC development across species, we sought to determine whether Hh signaling proteins are required for zebrafish PGC development, using both pharmacological and genetic perturbations of the Hh pathway. In this report, we demonstrate that cyclopamine, a small-molecule inhibitor of Smo, perturbs PGC motility in the zebrafish embryo, but Smo itself is dispensable for PGC homing to the presumptive gonad. Zygotic Smo activity also does not appear to be required for zebrafish PGC maturation or formation of the gonad itself, as both male and female zebrafish with reproductive organs derived from smu mutant germline tissues are fertile.
These observations raise the possibility that PGC migration has mechanistically diverged between invertebrates and vertebrates. Hh signaling mechanisms in Drosophila, zebrafish, and mice may also differ, since invertebrate- and vertebrate-specific pathway components have been identified (Varjosalo et al., 2006). Alternatively, the PGC migration defects previously attributed to altered Hh signaling in Drosophila could be due to coincident perturbations to other cellular processes. In support of this latter possibility, independent studies by Renault et al. (2009) argue against the involvement of Hh or its downstream effectors in Drosophila PGC migration. As described in the accompanying paper, Renault et al. (2009) did not observe PGC migration defects in embryos lacking maternal smo or ttv function, nor did they detect abnormal germ cell development when dominant negative or constitutively active versions of Hh signaling components were selectively expressed in these cells. They further observed that germ line clones mutant for ptc or pka failed to complete oogenesis, precluding any analysis of PGC migration in these genetic backgrounds, and they found that the chemoattractant activity of HMG-CoAR-expressing cells was Hh-independent, obviating any requirement for Hh signaling in directed PGC migration. Taken together, our results and those of Renault et al. (2009) call into question any role for Hh signaling in PGC migration.
Our observation that cyclopamine acts independently of Smo to disrupt zebrafish PGC migration is also surprising, as this small molecule has been used extensively to study Hh pathway-dependent patterning in this model organism. Since no other Smo homolog has been identified in zebrafish, cyclopamine most likely acts independently of the Hh pathway to induce PGC mislocalization. Our studies indicate that cyclopamine acts during the earliest stages of PGC migration (3 to 6 hpf), which is characterized by the cell-autonomous downregulation of E-cadherin levels (Blaser et al., 2005). One of the hallmarks of the onset of PGC migration is the cell-autonomous downregulation of E-cadherin levels, and our observations suggest that cyclopamine-induced PGC mislocalization is due at least in part to altered cell adhesive properties. PGCs in embryos treated with cyclopamine maintain cell-cell contacts for unusually long durations, and although our analyses focused on PGC-PGC interactions, it is likely that adhesive interactions between PGCs and somatic cells are similarly perturbed by cyclopamine. Indeed, our observation that cyclopamine act primarily through the soma to disrupt PGC migration suggests that increased cellular adhesion between PGCs and somatic cells and/or within the soma may be the dominant cause of cyclopamine-induced PGC mislocalization. Consistent with this model, PGC migration to the presumptive gonad site in cyclopamine-treated embryos can be partially rescued by globally reducing E-cadherin expression. Dysregulated PGC-soma adhesion could also explain why cyclopamine reduces the frequency of run phases during PGC movement and occasionally causes PGC fragmentation. These findings highlight the importance of properly regulated cell adhesion to permit cell migration, as has been demonstrated for border cells (Niewiadomska et al., 1999) and melanocytes (Nakagawa and Takeichi, 1998; Nishimura et al., 1999), among others.
How cyclopamine might dysregulate cell adhesive properties remains unclear. The timing of cyclopamine action also coincides with the onset of zygotic transcription in zebrafish embryos; however, microarray analyses do not reveal any significant changes in the expression of cell adhesion regulators upon cyclopamine treatment (data not shown). Cyclopamine might therefore perturb the activity of cell adhesion molecules in a post-transcriptional manner. Regardless of the precise mechanism of action, our observations urge caution in the interpretation of in vivo data obtained from cyclopamine-treated zebrafish embryos. For example, the effects of cyclopamine on early-stage embryos that have been ascribed to maternal Smo inhibition (Chen et al., 2001) may warrant genetic confirmation in MZsmuhi1640 zebrafish. Recent reports that cyclopamine blocks anchorage-dependent growth of breast cancer cells in a Smo-independent manner (Zhang et al., 2008) also raises the possibility that the cyclopamine-induced PGC defects we observe in zebrafish embryos may have implications for cellular behavior in other species.
Our observations shed some light on the mechanisms by which PGCs migrate. Cyclopamine-treated PGCs exhibit both a slower speed during run phases and a reduction in the fraction of time spent in these states, which together significantly abrogate cell migration. This motility defect contrasts phenotypes associated with loss of the chemoattractant sdf1a or its receptor cxcr4b, in which PGCs are observed less frequently in the run phase but migrate a normal speeds (Reichman-Fried et al., 2004). In addition, cyclopamine-treated PGCs demonstrate that cell polarization and translocation, which are intimately connected in the native condition, can be uncoupled.
Other features of PGC migration remain intact in cyclopamine-treated embryos. First, the fraction of time these cells exhibit an elongated morphology is unchanged since increased duration times compensate for reduced frequencies. This observation may reflect a fundamental property of PGCs: they are able to polarize and extend cellular protrusions even when motility is impaired. Indeed, random polarizations and protrusions are observed in PGCs prior to their acquisition of motility (Blaser et al., 2005). Second, PGCs in cyclopamine-treated embryos have run phase durations that resemble those of wildtype zebrafish, despite being inhibited in their frequency and speed, suggesting that this cellular behavior might be subject to an intrinsic clock mechanism. Elucidation of the molecular changes that occur upon cyclopamine exposure will provide further insights into the regulatory mechanisms of PGC migration.
We thank W. Talbot for syutbx392 zebrafish; R. Karlstrom for dtrts269 zebrafish; H. Knaut for anti-Vasa antibody; J. Marrs for anti-E-cadherin antibody; and B. Ciruna, I. Woods, M. Voas, M. Strasser, and B. Boldijapour for their technical assistance. This work was supported by an American Cancer Society Research Scholar Grant (RSG-08-041-01-DDC) to J.K.C., March of Dimes (1-FY2005-121) and National Institutes of Health (R01 HL069594) grants to D. Y., a California Institute for Regenerative Medicine Scholar Award (T1-0001) to J.K.M., and funding from the New York University Graduate Training Program in Developmental Genetics (NIH T32 HD007520) to N.A.T.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.