Abstract
Germ cells preserve an individual's genetic information and transmit it to the next generation. Early in development germ cells are set aside and undergo a specialized developmental programme, a hallmark of which is the migration from their site of origin to the future gonad1. In Drosophila, several factors have been identified that control germ-cell migration to their target tissues2, 3, 4; however, the germ-cell chemoattractant or its receptor have remained unknown. Here we apply genetics and in vivo imaging to show that odysseus, a zebrafish homologue of the G-protein-coupled chemokine receptor Cxcr4, is required specifically in germ cells for their chemotaxis. odysseus mutant germ cells are able to activate the migratory programme, but fail to undergo directed migration towards their target tissue, resulting in randomly dispersed germ cells. SDF-1, the presumptive cognate ligand for Cxcr4, shows a similar loss-of-function phenotype and can recruit germ cells to ectopic sites in the embryo, thus identifying a vertebrate ligand–receptor pair guiding migratory germ cells at all stages of migration towards their target.
In search of factors that affect the specification, maintenance or migration of primordial germ cells (PGCs) in zebrafish, we assayed embryos from a large-scale genetic screen for PGC number and position using vasa, a PGC-specific marker5, 6, 7. In zebrafish, PGCs originate at random positions with respect to the body axis. At 14 h after fertilization they align with the presomitic mesoderm resulting in one cluster on either side of the embryo. The PGCs then migrate caudally to the anterior yolk extension, the future site of the gonads (Fig. 1a, b). During migration, PGCs are repelled from entering the trunk by an unknown signal emanating from the pronephros, but they are attracted by the anterior yolk extension8, 9. Among 1,358 genomes screened, we identified one strictly zygotic mutation, odysseus (ody), that resulted in incorrect positioning of PGCs without affecting general embryo morphology. Instead of being localized to the gonadal anlage, PGCs in ody mutants fail to cluster and are found randomly dispersed throughout the body at 30 h after fertilization (Fig. 1c).
Figure 1: ody affects PGC positioning and is required within the germ line.
a–d, Whole-mount immunostaining with an antibody against Vasa5 labels PGCs (arrows) in wild-type (a, b) and ody mutant embryos (c, d). At 30 h after fertilization, PGCs cluster at the anterior yolk extension in wild-type embryos (a), whereas in ody mutants (c), they are dispersed throughout the body. At 9 days after fertilization, PGCs have coalesced with the gonadal tissue in wild type (b), whereas in ody mutants no Vasa-positive PGCs are detectable (d). e–j, Transplantation of wild-type and ody mutant PGCs into wild-type and ody mutant embryos. One-cell-stage donor embryos were injected with biotin-dextran, and at the 8,000-cell stage approximately 100 donor cells were transplanted into recipient embryos of an equivalent stage. PGCs were identified by Vasa antibody staining (green) and donor-derived cells by anti-biotin antibody labelling (red). Transplantation of wild-type PGCs into wild-type embryos (e, magnified view in h) does not affect their ability to reach the anterior yolk extension. ody mutant PGCs transplanted into wild-type embryos (g, magnified view in j) still go astray, whereas wild-type PGCs transplanted into ody mutant embryos (f, magnified view in i) arrive at their target. Arrows indicate donor-derived PGCs; arrowheads indicate recipient-derived PGCs.
High resolution image and legend (52K)To determine whether ody acts within the PGC or in the somatic tissue we generated genetic mosaics. On transplantation into ody mutant embryos, wild-type PGCs migrate to the correct position (n = 9, Fig. 1f, i). However, mutant PGCs transplanted into wild-type embryos do not reach their target (n = 3, Fig. 1g, j). This indicates that the ody gene product is required within the PGCs for proper migration, in contrast to previously identified factors controlling PGC migration that act exclusively in somatic tissues2, 3, 4.
The random positioning of PGCs in ody mutants could simply be the result of generally immotile cells being unable to leave their randomized sites of origin. Alternatively, ody might specifically affect the response of PGCs to chemotactic signals. In vivo imaging of embryos using a PGC-specific green fluorescent protein (GFP) marker9 revealed that mutant PGCs (Fig. 2e–g; see also Supplementary movie 2) display a similar degree of overall motility to wild-type cells (Fig. 2 a–c; see also Supplementary movie 1). However, ody mutant PGCs fail to respond to both attractive and repulsive regions, and individual cells show faulty migration to various locations in the embryo, indicative of defective chemotaxis.
Figure 2: PGC migration in wild-type and ody mutant embryos.
Expression of GFP9 (a–c, e–g) or GFP fused to a PH domain12,13 (d, h) was targeted to PGCs by fusing the coding sequences to the zebrafish vasa 3' untranslated region (UTR)16. Capturing consecutive images of individual embryos shows that PGCs (arrows indicate initial PGC position) cluster and migrate in a posterior direction in wild-type embryos (a–c, anterior is to the top right). In ody mutant embryos (e–g, anterior is to the bottom left) PGCs (arrows) do not cluster, and migrate individually to ectopic sites. Germ cells recruit PH–GFP to locally restricted sites at the membrane (arrows, d, h) both in wild-type and ody mutants.
High resolution image and legend (31K)In many migratory cells, chemoattractants are sensed by G-protein-coupled receptors that signal through phosphatidylinositol-3-OH kinase (PI(3)K) to recruit pleckstrin homology (PH) domain-containing proteins to the leading edge (reviewed in refs 10, 11). Therefore, we investigated the subcellular localization of a PH–GFP fusion protein12, 13 in PGCs in vivo. In wild-type PGCs, membrane recruitment of PH–GFP is locally restricted to the site of lamellipodium protrusion and remains relatively stably positioned over a longer period of time (Fig. 2d; see also Supplementary movie 3). Similarly, ody mutant PGCs recruit PH–GFP locally to the leading edge, but the site of recruitment changes position frequently (Fig. 2h; see also Supplementary movie 4). This indicates that ody mutant PGCs possess a functional machinery for establishing localized PI(3)K signalling and execution of the downstream migratory programme; however, they fail to maintain stably localized PI(3)K activation and thus lose directionality. This suggests that ody mutant PGCs do not perceive guidance cues. In mouse and zebrafish, PGCs migrate in close association1 and ody seems to be required for this clustering in zebrafish. Although clustering might be a secondary effect of individual PGCs migrating to a common target, it is difficult to see how this might promote the intimate cell–cell contact observed between individual wild-type PGCs.
In principle, the ody mutation might affect PGC migration only indirectly by interfering with the normal PGC developmental programme. We find, however, that in 46% (n = 78) of the homozygous ody mutant embryos the gonads are partially populated with one to four PGCs and that a similar fraction of homozygous adult fish are fertile. Thus, ody mutant PGCs are, apart from the migration defect, normal with respect to their developmental programme and function.
We mapped ody (2,000 meioses) to a genomic interval of less than 200 kilobases (kb), 50.9 cM from the top of linkage group 9. Synteny and genome sequence analysis yielded two candidate genes, arhgef4 and cxcr4b. Sequencing demonstrated that the G-protein-coupled receptor cxcr4b bears a nonsense mutation (Lys 239 to stop), deleting the carboxy terminal part of the third intracellular loop and the last two transmembrane domains. As these domains have been shown to be required generally for G-protein-coupled receptor signalling14, this mutation probably results in a complete loss of Cxcr4b protein function. cxcr4b is expressed in various tissues15 and in situ hybridization shows that it is also expressed in migrating PGCs (Fig. 3a, b). Targeting the expression of wild-type cxcr4b to PGCs16 fully rescues the ody mutant and injection of cxcr4b antisense morpholino oligonucleotides reproduces the ody mutant phenotype (Fig. 3c, d), showing that cxcr4b is the gene compromised in ody mutants. cxcr4b belongs to a subfamily of G-protein-coupled chemokine receptors, members of which have, among other processes, been implicated in leukocyte migration17. Human CXCR4, known as fusin, forms, together with CD4, a receptor complex for HIV-1 (ref. 18). Mutant mice devoid of Cxcr4 function are defective in B-cell lymphopoiesis, neuronal migration and vascularization19, 20. The comparatively subtle phenotype of ody/cxcr4b mutants may be due to the presence of two Cxcr4 receptors with potentially split function in zebrafish, such that disruption of cxcr4b reveals only a subset of the defects seen in mouse. This is supported by their exclusive expression patterns15. Alternatively, there may be variation between zebrafish and mouse in the roles of the Cxcr4 receptors.
Figure 3: The G-protein-coupled chemokine receptor Cxcr4b is required for PGC guidance in zebrafish.
a, b, In situ hybridization reveals that at the seven-somite stage Cxcr4b is, among other tissues15, expressed in two lateral clusters at the level of the second somite (arrows in a; the boxed area in a is magnified in b). Co-staining for Cxcr4b messenger RNA (green) and Vasa protein (red) shows that these two lateral clusters correspond to PGCs (b). c, Embryos injected at the one-cell stage with Cxcr4b anti-sense morpholinos (0.2 mM, 5'-AGTGTGCTCAAAAAGGCGCAATAAG-3') display scattered PGCs (revealed by anti-Vasa staining) at 30 h after fertilization and thus recapitulate the ody phenotype. d, Injection of mismatch Cxcr4b morpholinos (0.2 mM, 5'-AGAGTCCTGAAAAAGGCGGAAAAAG-3') has no effect on PGC migration.
High resolution image and legend (33K)As cell culture experiments and loss of function analysis in mouse have implicated the chemokine SDF-1 as the ligand for Cxcr4 (refs 21–23), we analysed a zebrafish homologue of SDF-1. SDF-1 is expressed in a dynamic pattern that prefigures the route of PGC migration (Fig. 4a–d). Reducing SDF-1 activity by antisense morpholinos results in incorrect positioning of PGCs as in ody mutant embryos (28 out of 85) (Fig. 4e, g). Moreover, ectopic expression of SDF-1 in wild-type embryos recruited between 10% and 30% of the endogenous PGCs to positions near the SDF-1-expressing cells to which they would never normally migrate (15 out of 23) (Fig. 4f). This response is receptor-dependent as SDF-1-overexpressing cells failed to attract ody mutant PGCs to ectopic positions (0 out of 15) (Fig. 4h).
Figure 4: SDF-1 is a chemoattractant for PGCs.
Zebrafish SDF-1 RNA is expressed in a dynamic pattern that demarcates the prospective PGC migration route (a–d, g). a, c, In a dorsal view, in situ hybridization with a SDF-1 probe (a) and immunostaining against Vasa protein (c) show that SDF-1 is expressed proximal to the two PGC clusters (arrows) at the level of the second somite in a seven-somite-stage embryo. b, d, In a 17-somite-stage embryo, the anterior boundary of the SDF-1 expression domain (b, arrow) reaches to the level of the eighth somite where the PGC cluster resides (d, arrow). g, At 36 h after fertilization the SDF-1 expression domain is restricted to the final target of PGCs (arrows). e, Embryos injected with SDF-1 anti-sense morpholinos (2 mM, 5'-CGCTACTACTTTGCTATCCATGCCA-3') display scattered PGCs at 30 h after fertilization, a phenotype indistinguishable from the ody mutant (compare with Fig. 1c). SDF-1 mismatch morpholinos (2 mM, 5'CGGTATTACTTTCCTATCCTTGGCA3') do not affect PGC migration (data not shown). f, Transplantation of blastula cells from embryos injected with in vitro-transcribed SDF-1 mRNA and biotin-dextran into wild-type embryos attracts PGCs (green) towards SDF-1-expressing cells (red). h, Transplanting similarly treated cells into ody mutant embryos has no effect on PGC distribution. Arrows indicate incorrectly positioned PGCs; arrowheads indicate correctly positioned PGCs (f, h).
High resolution image and legend (61K)From these results we conclude that the attractant SDF-1 and its receptor ody/Cxcr4b comprise a system for guiding vertebrate PGCs throughout their entire migration. Furthermore, this finding reveals a role for G-protein-coupled receptors in embryonic development—other roles include the mediation of attractive chemotaxis of Dictyostelium amoebae, leukocytes10, 11 and, more recently, neurons24. Notably, PGCs with inactive SDF-1 or cxcr4b also ignore repulsive signals. Although this might simply be due to the fact that the early faulty migration of PGCs means that they never encounter such signals, an alternative explanation would be that SDF-1 and cxcr4b are also mediating repulsion, either by a repellent directly interfering with the system or by attractant-activated processes being a prerequisite for responding to repulsion. Given the length and complexity of the migration route in zebrafish, it seems unlikely that PGC guidance relies solely on the SDF-1/Cxcr4b pair. When considering the multiple sites of SDF-1 expression, it is likely that either post-transcriptional modifications of SDF-1 or co-expression of repellents may assist PGC navigation. In Drosophila, two enzymes involved in lipid metabolism, Wunen and Columbus2, 3, 4, affect PGC migration presumably by controlling production of unknown repulsive or attractive signals. It will be interesting to see how such signals might tie in with the guidance system identified here and whether a widely conserved system exists for guiding PGCs to their gonadal targets.
The Tübingen 2000 Screen Consortium
E. Busch-Nentwich*, R. Dahm*, H.-G. Frohnhöfer*, H. Geiger*, D. Gilmour*, S. Holley*, J. Hooge*, D. Jülich*, H. Knaut*, F. Maderspacher*, H.-M. Maischein*, C. Neumann*, T. Nicolson*, C. Nüsslein-Volhard*, H. Roehl*, U. Schönberger*, C. Seiler*, C. Söllner*, M. Sonawane*, F. van Bebber*, A. Wehner*, C. Weiler*, P. Erker†, H. Habeck†, U. Hagner†, C. E. Hennen Kaps†, A. Kirchner†, T. Koblizek†, U. Langheinrich†, C. Loeschke†, C. Metzger†, R. Nordin†, J. Odenthal†, M. Pezzuti†, K. Schlombs†, J. deSantana-Stamm†, T. Trowe†, G. Vacun†, B. Walderich†, A. Walker†, C. Weiler†
* Max Planck Institut für Entwicklungsbiologie, Abteilung III/Genetik, † Artemis Pharmaceuticals GmbH, an Exelixis company, Spemannstrasse 35, 72076 Tübingen, Germany
