Development 126, 5295-5307 (1999) 5295 Printed in Great Britain © The Company of Biologists Limited 1999 DEV3057

Identification of tissues and patterning events required for distinct steps in early migration of zebrafish primordial germ cells

Gilbert Weidinger‡, Uta Wolke‡, Marion Köprunner, Michael Klinger* and Erez Raz¦ Department of Developmental Biology, Institute of Biology I, University of Freiburg, Hauptstrasse 1, D-79104 Freiburg, Germany *Present address: Faculty of Biology, University of Konstanz, Universitätsstrasse 10, D-78434 Konstanz, Germany ‡These authors contributed equally to this work ¦Author for correspondence (e-mail: [email protected])

Accepted 21 September; published on WWW 9 November 1999

SUMMARY

In many organisms, the primordial germ cells have to obtain information on the origin of the positional cues migrate from the position where they are specified towards provided to the germ cells by somatic tissues during their the developing gonad where they generate gametes. migration, we analyzed the migration pattern in mutants, Extensive studies of the migration of primordial germ cells including spadetail, swirl, chordino, floating head, cloche, in Drosophila, mouse, chick and Xenopus have identified knypek and no isthmus. In mutants with defects in axial somatic tissues important for this process and structures, paraxial mesoderm or dorsoventral patterning, demonstrated a role for specific molecules in directing the we find that certain steps of the migration process are cells towards their target. In zebrafish, a unique situation specifically affected. We show that the paraxial mesoderm is found in that the primordial germ cells, as marked by is important for providing proper anteroposterior expression of vasa mRNA, are specified in random positions information to the migrating primordial germ cells and relative to the future embryonic axis. Hence, the migrating that these cells can respond to changes in the global cells have to navigate towards their destination from dorsoventral coordinates. In certain mutants, we observe various starting positions that differ among individual accumulation of ectopic cells in different regions of the embryos. Here, we present a detailed description of the embryo. These ectopic cells can retain both morphological migration of the primordial germ cells during the first 24 and molecular characteristics of primordial germ cells, hours of wild-type zebrafish embryonic development. We suggesting that, in zebrafish at the early stages tested, the define six distinct steps of migration bringing the vasa-expressing cells are committed to the germ cell primordial germ cells from their random positions before lineage. gastrulation to form two cell clusters on either side of the midline by the end of the first day of development. To Key words: Zebrafish, Primordial germ cell, Cell migration, vasa

INTRODUCTION cells along the migratory route. Receptor-ligand interaction is required to support migration and survival of the PGCs While the specification mechanisms of primordial germ cells (Bernex et al., 1996; Matsui et al., 1990). Mouse PGCs have (PGCs) and the position where these cells are specified can also been shown to migrate towards explants of target tissue, differ, a common theme for many species is that the germ suggesting that they are attracted towards the genital ridge by cells are formed in regions distinct from the site where the long-range signaling (Godin et al., 1990). Finally, a role for gonad will form. Hence, the PGCs have to migrate towards extracellular matrix (ECM) molecules in the PGC migration the future gonad, a process that has been studied in Xenopus process has been suggested by the finding of specific and particularly in Drosophila and mouse where modern interactions between ECM molecules and PGCs and by the genetic approaches have been applied (reviewed in Howard, phenotype of PGCs lacking specific receptors for components 1998; Rongo et al., 1997; Wylie, 1999). The general of the ECM (e.g., Anderson et al., 1999). In Drosophila, a conclusion from these studies is that the movements of the detailed description of the migration process coupled with PGCs towards the gonadal region, where they associate with genetic analysis allowed the identification of tissues and cells of mesodermal origin, rely on directional cues provided molecules required for PGC migration (reviewed in Howard, by the somatic environment (e.g., Anderson et al., 1999; 1998; Rongo et al., 1997; Williamson and Lehmann, 1996; Jaglarz and Howard, 1994; Matsui et al., 1990; Moore et al., Wylie, 1999). Drosophila PGCs are formed at the posterior 1998; Zhang et al., 1996). A classical example from the pole of the early embryo and then together with somatic cells mouse is that of the c-kit receptor, which is expressed in the participate in the morphogenetic movement forming the PGCs, and its ligand Steel, which is expressed by somatic posterior midgut (PMG). Subsequently, the PGCs actively 5296 G. Weidinger and others migrate through the midgut epithelium towards the gonadal MATERIALS AND METHODS mesoderm with which they align and coalesce. Two genes specifically affecting this migration are expressed in somatic Zebrafish maintenance and mutant strains cells (Moore et al., 1998; van Doren et al., 1998; Zhang et Zebrafish (Danio rerio) were maintained as described previously al., 1996, 1997): the wunen gene is required for repulsion of (Westerfield, 1995). Mutant strains used: chordino, dintm84 PGCs, thereby bringing them closer to the mesodermal target (Hammerschmidt et al., 1996); cloche, clom39 (Stainier et al., 1995); whereas columbus functions in attracting them towards the floating-head, flhtk241 (Odenthal et al., 1996); knypek, knym818 (A. Chitnis, K. Artinger and W. Driever, personal communication; gonadal mesoderm. tu29a Until recently, studying the migration of PGCs in fish has Solnica-Krezel et al., 1996); no isthmus, noi (Brand et al., 1996); spadetail, sptb104 (Kimmel et al., 1989); swirl, swrta72a (Mullins et al., relied on their identification by morphological criteria at 1996). relatively late stages, usually around the beginning of somitogenesis (e.g., Gevers et al., 1992). The ability to follow Whole-mount in situ hybridization and histology PGC migration in fish was revolutionized by the cloning of the Two-color in situ hybridization was performed as described by Jowett zebrafish vasa gene homolog, which is exclusively expressed and Lettice (1994) with modifications according to Hauptmann and in PGCs (Olsen et al., 1997; Yoon et al., 1997). vasa was Gerster (1994). In some cases, light-blue color was obtained using a originally identified in Drosophila as a maternal effect gene β-galactosidase-conjugated anti-digoxigenin and X-Gal for required for the formation of the abdominal segments and for subsequent color reaction (Hauptmann, 1999). The following probes germ cell specification (Hay et al., 1988; Lasko and Ashburner, were used for whole-mount in situ mRNA hybridization: din (Miller- 1988; Schüpbach and Wieschaus, 1986). In situ hybridization Bertoglio et al., 1997), gata2 (Detrich et al., 1995), hoxa-2 (Prince et of 4-cell-stage zebrafish embryos showed specific localization al., 1998), myoD (Weinberg et al., 1996), papc (Yamamoto et al., 1998), pax2.1 (formerly paxb, (Krauss et al., 1991)), pax8 (Pfeffer et of the vasa transcript in four stripes at the edges of the first two al., 1998) and vasa (Olsen et al., 1997; Yoon et al., 1997). cleavage planes, suggesting that specification of germ cells is Methacrylate sections were performed using the JB-4 Plus resin achieved by localization of cytoplasmic components as in (Polyscience Inc.) according to the manufacturer’s protocol. other organisms such as Drosophila, C. elegans and Xenopus (Wylie, 1999; Yoon et al., 1997). At the 32-cell stage, the Determination of PGC number transcript is detected in four cells that subsequently divide to vasa-positive cells were counted using a stereomicroscope at ×150 give rise to four cell clusters located close to the blastoderm magnification. The number of vasa-positive PGCs varies among margin at late blastula stages. During subsequent development embryos of the same clutch. Importantly, the number of PGCs may these vasa-expressing cells migrate and at 24 hours also be influenced by the genetic background. For example, mutant postfertilization (hpf) end up in two bilateral rows around the chordino and wild-type siblings in the TL genetic background have anterior end of the yolk extension. These cells that maintain about 25 PGCs at 24 hpf (e.g., Table 1), while embryos with AB genetic background typically have as many as 50 PGCs at 24 hpf (e.g., vasa expression at least up to late larval stages are primordial Table 2). For these reasons, description of mutant phenotypes was done germ cells based on their morphology and position (Yoon et by comparing the PGC number or behavior to wild-type siblings. al., 1997). Additionally, to minimize the variability, analysis of the migration Here we provide a detailed analysis of the early stages of process in wild-type and in mutant embryos was usually performed by PGC migration in zebrafish. We find that this process can be fixing embryos of the same clutch at different times of development. divided into six distinct steps, some of which appear to be shared with somatic tissues, whereas others seem to reflect vasa cDNA amplification and injection of mRNA active migration of PGCs relative to their somatic neighbors. Full-length vasa cDNA (as published by Yoon et al. (1997)) was Analysis of the migration process in mutant embryos shows amplified from ovary cDNA using the following primers: 5′ primer that some of these steps can be specifically affected. This (TCA GGC TCT TCA CGC GTG TCC ACC TGC TAC), 3′ primer allows us to propose which structures or processes are (TTT TGT CAC CAG TAT CCG TCT TTA TTT TGA) (italic sequence involved in directing the migration of the PGCs. As a result is homologous to vasa). Except for a few amino acid changes that we attribute to polymorphisms between strains (identical changes were of abnormal migration, in some mutants we observe detected in independent PCR reactions), the sequence of the cloned accumulation of cells in ectopic regions. As judged by vasa cDNA is identical to that published by Yoon et al. (1997). morphological and molecular criteria, these ectopic PGCs can Capped vasa mRNA was synthesized using the Ambion Message maintain their predetermined fate in a foreign somatic Machine kit and zebrafish embryos were injected at the 1- to 2-cell environment at the early stages tested. stage with 100-200 pg of vasa mRNA. In situ hybridization was

Table 1. PGC phenotype of chordino (din) mutant embryos Average PGC Embryos with PGCs Proportion of PGCs Embryos analyzed number per embryo* ventral-posterior of main clusters ventral-posterior of main clusters Stage wt siblings din wt siblings din wt siblings din wt siblings din 60% epiboly 13 15 23.1±5.7 22.0±4.6 100% 100% 60.0%±11.4% 63.4%±13.7% 80% epiboly 14 14 21.9±4.8 21.6±4.3 100% 100% 58.9%±13.7% 56.0%±13.1% 2 somite 24 9 24.6±7.3 23.3±6.9 92% 100% 19.4%±13.1% 34.7%±19.3% 8 somites 19 14 33.7±8.6 34.4±8.4 100% 100% 21.8%±9.6% 41.8%±13.6% 24 hpf 52 55 26.4±7.4 27.4±8.9 15% 95% 1.5%±4.2% 27.4%±14.4%

* in TL genetic background (see Materials and Methods). Primordial germ cell migration in zebrafish 5297

Table 2. PGC phenotype of spadetail (spt) mutant embryos Average PGC Embryos with PGCs Proportion of PGCs Embryos analyzed number per embryo* anterior of main clusters anterior of main clusters Stage wt siblings spt wt siblings spt wt siblings spt wt siblings spt 1 somite 13 10 50.4±9.7 56.2±10.8 46% 100% 2.4%±3.3% 35.9%±15.0% 5 somites 29 25 39.1±6.4 39.7±12.6 48% 93% 3.7%±5.0% 21.3%±13.2% 24 hpf 85 53 45.6±9.6 47.4±8.4 17% 96% 1.0%±2.6% 16.5%±10.6% 48 hpf 8 10 n.d. 49.8±12.1 0% 100% 0% 26.9%±10.3%

Embryos with PGCs Proportion of PGCs posterior of main clusters posterior of main clusters Stage wt siblings spt wt siblings spt 1 somite 100% 100% 17.7%±10.0% 21.6%±8.1% 5 somites 97% 100% 22.3%±13.8% 18.6%±12.2% 24 hpf 20% 72% 1.1%±3.0% 8.7%±9.7% 48 hpf 0% 75% 0% 9.1%±9.8%

n.d., not done * in AB genetic background (see Materials and Methods). performed on the injected embryos at different stages. Judged from embryos, proving that the arrangement of the clusters is the signal intensity at early gastrulation, most cells in the injected random relative to the dorsal aspect of the embryo (data not embryos received higher levels of vasa mRNA than the endogenous shown). Three examples for initial cluster arrangements are level of vasa expressed by the PGCs. shown in Fig. 1. The dorsal side of the embryo can be flanked by two PGC clusters (starting position A, Fig. 1A), in an intermediate arrangement one of the clusters is located off the RESULTS middle of the chordino expression domain either on the left or the right side (starting position B, Fig. 1B), or one of the Migration of primordial germ cells in wild-type clusters lies exactly at or very close to the dorsal side (starting embryos position C, Fig. 1C). Therefore, in contrast to other organisms, To identify possible sources for signals and define tissues that within each zebrafish embryo, the PGCs start their migration may participate in directing PGCs towards their target, we from different dorsoventral positions and these starting points analyzed the migration of PGCs relative to forming somatic differ between individual embryos. Below we describe the structures in the first 24 hours of development. As discussed movements that take place during the next 20 hours of by Yoon et al. (1997), based on their morphology and position development bringing the PGCs to their position at the 1-day at larval stages, the vasa-expressing cells are PGCs. It is stage. These movements can be divided into six steps, some of formally possible that some cells do not maintain expression which are temporally distinct, while others occur of vasa or that not all of the vasa-positive cells at a certain simultaneously. stage were expressing it at earlier stages. Our results (see below) do not support these options, but rather are consistent Step I, convergence towards the dorsal with the notion that, during the first 48 hpf, the vasa mRNA is At the 60% epiboly stage (6.5 hpf, early gastrulation), we still a reliable, stable marker for PGCs and that cells expressing it find three classes of PGC arrangements. The positions of PGC are related by lineage. Throughout this work then, the analysis clusters and the frequencies with which each arrangement is of the migration of the vasa-expressing cells during the first 24 found (data not shown) imply that they derive from the three hours of development was performed by observing the position arrangements described for the dome stage; however, at 60% of these cells at different stages and deducing movements that epiboly, the PGC clusters are located more dorsally (Fig. 1A- would connect the different arrangements observed at different C, middle panel). This movement starts relatively early, before time points. the onset of gastrulation, since at shield stage (6 hpf, onset of Observing the migration of the vasa-expressing PGCs is gastrulation), the square formed by the PGC clusters is already complicated by the fact that the position of the four PGC lost in most embryos (Fig. 2A). This first step of PGC clusters at blastula stages seems to be determined by the migration appears to be shared with somatic cells that at this orientation of the early cleavage planes (Yoon et al., 1997), stage undergo compaction (Warga and Nüsslein-Volhard, which is random with respect to the future dorsoventral axis 1998) and convergence movements (Solnica-Krezel et al., (Abdelilah et al., 1994; Helde et al., 1994). To prove that the 1995) towards the dorsal. initial relation of the four PGC clusters relative to the dorsal aspect of the embryo is indeed random, we analyzed embryos Step II, exclusion from the dorsal midline before gastrulation. At dome stage (late blastula, 4.5 hpf), the As the embryonic shield (the zebrafish organizer) is formed, four PGC clusters, each containing about four cells, are found PGC clusters located very close to the dorsal (starting position close to the blastoderm margin, equidistant from each other in C) are excluded from the midline to occupy a slightly more a square-like arrangement (Fig. 1A-C, upper panel). We lateral position while more lateral clusters continue to converge determined the orientation of this square relative to the dorsal dorsally. In general, no germ cells are observed in the extreme part of the embryo (visualized by chordino expression) in 30 dorsal region by 60% epiboly (Fig. 1C, middle panel). 5298 G. Weidinger and others

Fig. 1. Migration of PGCs in wild-type embryos. The PGCs are labeled using the vasa mRNA probe (dark blue) and other structures are labeled in red or in light blue with the probes indicated. Embryos at somitogenesis stages were deyolked and flattened. (A-C) The three basic initial PGC arrangements relative to the dorsal aspect of the embryo are shown at the dome stage (upper panel), where dorsal is marked by the chordino expression domain. At 60% epiboly (middle panel) and the 2-somite stage (lower panel) also three classes of arrangements are found at frequencies expected from the three initial arrangements at the dome stage. The 60% epiboly stages are stained with chordino (PGC clusters are marked with arrowheads), the 2-somite stages with myoD marking the adaxial cells, papc expressed in the segmental plate and the forming somites, and pax8 staining the otic placodes. Dome and 60% epiboly stages are shown in animal view with dorsal up. (D,E) PGCs align along the border of the trunk mesoderm that expresses papc both on the dorsal (D) and the ventral (E) side at the 90% epiboly stage. Note that no PGCs are found on the notochord, which is devoid of papc staining in D. (F) A 1-somite- stage embryo showing the alignment of posterior PGCs (arrowheads) at the lateral border of the broad expression domain of pax8 in the pronephric anlage. Pax8 also stains the otic placodes and myoD the adaxial cells (dark blue). (G,H) Cross-sections of 4-somite-stage embryos after whole-mount in situ hybridization with vasa. (G) At the level of somite 1, two medially located PGCs are seen in contact with the YSL and the overlying paraxial mesoderm (arrowheads), whereas two laterally located PGCs have lost contact with the YSL and extend up to the ectoderm (arrows). (H) A posterior trailing PGC (arrow) at the level of somite 7 found at the lateral margin of the mesoderm in contact with the YSL and the ectoderm. (I) 8-somite-stage embryo stained with myoD (adaxial cells and somites) and pax2.1 (pronephros, otic placodes, midbrain-hindbrain boundary and eye anlagen). There is one ectopic anterior PGC present in this wild-type embryo (arrow). (J) A 16- somite-stage embryo stained with myoD. Note that the PGC clusters have shifted towards the posterior, whereas trailing cells have migrated anteriorly. (K,L) PGCs are located in two lateral lines at the anterior end of the yolk extension at 24 hpf as seen in lateral (K) and dorsal (L) view.

Step IIIa, alignment along the anterior border of the is shown in Fig. 1D for the 90% epiboly stage. As development trunk mesoderm proceeds, this pattern becomes progressively more defined as PGC clusters initially positioned close to the dorsal side align the PGCs form a 1- to 3-cell-wide line on either side of the along the boundary between trunk and head paraxial mesoderm axis at the anteroposterior level of the first somite (Fig. 1A-C, marked by the anterior border of the paraxial protocadherin lower panel). Cross sections of this region at the 1- and 4- (papc) (Yamamoto et al., 1998) expression domain. The somite stage reveal that the aligned PGCs are always in close alignment can be first observed at the 60% epiboly stage and contact with the yolk syncytial layer (YSL) and the overlaying Primordial germ cell migration in zebrafish 5299

100 Step IIIb, alignment along the lateral border of the A mesoderm 80 square PGC clusters located more ventrally also align along the border of the papc expression domain from the 60% epiboly stage on 60 converged (shown in Fig. 1E at 90% epiboly). Later these PGCs remain 40 aligned to the outer border of the ventral mesoderm that starts to express pax8 in the anlage of the pronephros (Pfeffer et al., 20 other / 1998) (Fig. 1F, arrowheads). Cross sections at the 4-somite

Number of embryos (%) irregular stage show that these posterior trailing PGCs are in close 0 contact with the YSL and extend to the ectoderm (Fig. 1H). dome shield 60% epiboly At the 2-somite stage, the anterior and the lateral alignment (n=28) (n=31) (n=30) of PGCs (step IIIa and IIIb) are clearly visible and, depending 100 on the initial orientation of the clusters, three basic arrangements can be identified (Fig. 1A-C, lower panel). B A 80 D Step IV, formation of two lateral PGC clusters 60 Between the 1- and the 5-somite stages, the rows of PGCs at the level of the 1st somite migrate away from the axis and form 40 V two clusters lateral to the paraxial mesoderm extending from the 1st to the 3rd somite level (compare Fig. 1A-C, lower panel P Number of PGCs (%) 20 with Fig. 1I). Interestingly, as the PGCs form these clusters, many cells appear to dissociate from the YSL (Fig. 1G, 0 arrows). 95% epiboly (n=889) 2 somites (n=1056) 8 somites (n=1165) Step V, anterior migration of trailing PGCs 100 From the 95% epiboly stage to the 24 hpf stage, the percentage 80 as well as the absolute number of PGCs in regions posterior of the main cluster progressively decreases (Fig. 2B and data not 60 shown). Thus, from late gastrulation stages on, PGCs that have aligned along the lateral mesoderm border in posterior regions 40 migrate towards the anterior along the pronephros and join the main PGC clusters (for example, compare the position of these

Number of PGCs (%) 20 trailing cells in Fig. 1I and J). The formal possibility that this phenomenon results from 0 selective death of posteriorly positioned PGCs is not supported 16 somites (n=1277) 19 somites (n=1349) 24 hpf (n=1216) by our analysis of embryos with abnormal PGC migration patterns. PGCs expressing the vasa marker can survive for Fig. 2. In wild-type embryos, the PGCs converge towards the dorsal during gastrulation and migrate anteriorly during somitogenesis extended periods of time in various ectopic locations including stages. (A) The frequency of embryos showing the symmetrical posterior positions along their normal route (e.g., in chordino ‘square’ shape arrangement of the four PGC clusters decreases from mutant embryos, see below). the dome stage to the 60% epiboly stage while more embryos show the ‘converged’ arrangement where clusters are shifted towards the Step VI, posterior positioning of the PGC clusters dorsal. Irregular arrangements (3 or 5 clusters) appear at a constant The two PGC clusters initially form during step IV at the level frequency of around 10%. n is the number of embryos analyzed. of the 1st to 3rd somite. This position shifts towards the (B) Distribution of PGCs from late gastrula (95% epiboly) to 24 hpf. posterior so that, at the 16-somite stage, the clusters are located The number of vasa-positive cells was counted in four regions as at the level of the 5th to the 7th somite and by 24 hpf around indicated in the drawings of embryos at each stage. For the 95% somite 8 along the anterior part of the yolk extension (Fig. 1J- epiboly stage, the regions were sectors along the dorsoventral axis as seen from the vegetal pole. For the 2- to 16-somite stages, an L). illustration of a flattened embryo is shown with the anterior up; the A summary of these six steps of early PGC migration that regions were defined relative to tissues stained by myoD, papc and bring the PGCs from their initial random distribution into the pax8 for 2 somites (see Fig. 1A-C, lower panel), myoD and pax2.1 two clusters on either side of the axis is presented in Fig. 3. for 8 somites (see Fig. 1I) and myoD for 16 somites (see Fig. 1J). For the 19-somite and 24 hpf stages, the regions were defined relative to Migration of PGCs in mutants affected in structures like the yolk extension. Note that for all stages the blue dorsoventral patterning region contains the main cluster of PGCs, whereas the red and As a first test for the control of somatic tissue over PGC yellow regions are of the same absolute size. n is the total number of migration, we followed the migration pattern in mutants PGCs counted; for each stage 26 to 31 embryos were analyzed. affecting the global dorsoventral patterning of the embryo. Genes affecting this process have been identified in zebrafish and, as in Xenopus, a balance between ventralizing bone paraxial mesoderm (for the 4-somite stage see Fig. 1G, morphogenetic proteins (BMPs) and their antagonists secreted arrowheads). by the organizer has been shown to pattern the embryo along 5300 G. Weidinger and others A B C I I I G I I I a dorsal dorsal I I I b dorsal

dome, animal view 60% epiboly, animal view 80% epiboly, side view D E F

anterior anterior I V V I

V

2 somites 8 somites 19 somites 24 hpf

Fig. 3. The six steps of early PGC migration in zebrafish. Schematic drawings of embryos from dome stage to 24 hpf showing the positions and movements of the four PGC clusters for starting position B (see Fig. 1B). At dome stage, four clusters of PGCs are found close to the blastoderm margin in a symmetrical ‘square’ shape. All possible orientations of the square relative to the dorsal side of the embryo can be observed. (A) Here, an intermediate arrangement is shown with one cluster close to, but not directly at the dorsal side (see Fig. 1B). Beginning before gastrulation, lateral and ventral clusters move towards the dorsal, with ventral clusters migrating more slowly (step I, convergence towards the dorsal). This movement is shared with somatic cells and can be attributed to early compaction before gastrulation and dorsal convergence of hypoblast cells during gastrulation. (B) Clusters located very close to the dorsal migrate away from the dorsal midline and are therefore rarely found on the notochord from the 60% epiboly stage on (step II, exclusion from the dorsal midline). At the 60% epiboly stage, these movements have resulted in loss of the ‘square’ arrangement of PGC clusters; convergence continues, but some PGCs can still be found in far ventral positions until the end of gastrulation. (C) Dorsally located PGCs align along the border between the head and trunk paraxial mesoderm marked by the anterior border of the papc expression domain depicted by a dashed line (step IIIa, alignment along the anterior border of the trunk mesoderm). Ventrally located clusters align at the lateral border of the mesoderm that early on is also marked by the border of papc expression and at late gastrulation starts to express the pronephros marker pax8 (step IIIb, alignment along the lateral border of the mesoderm). (D) At the 2-somite stage, most PGCs have arrived in two lines at the level of the first somite. These anterior located PGCs migrate towards the lateral (step IV, formation of two lateral PGC clusters). Cells that were initially located ventrally migrate towards the anterior along the anlage of the pronephros (step V, anterior migration of trailing PGCs). In this illustration, the positions of the PGCs are drawn relative to the adaxial cells, the somites and the lateral border of the pronephric anlage. (E) At the 8-somite stage, all anterior PGCs are found lateral to the paraxial mesoderm in a cluster extending from the 1st to the 3rd somite. These clusters start to move towards the posterior (step VI, posterior positioning of the PGC clusters), while the trailing cells tightly align on the lateral border of the pronephros and continue to migrate anteriorly. Here, the PGCs are drawn relative to the expression domains of myoD in the adaxial cells and somites and pax2.1 in the pronephros (see Fig. 1I). (F) At the 19-somite stage, the main clusters have shifted to more posterior positions and in 60% of embryos some trailing cells are still seen. (G) At 24 hpf, the PGC clusters are located at the anterior end of the yolk extension, which corresponds to the 8th to 10th somite level. In most embryos, all PGCs have reached this region, only a few trailing cells are found close to the main clusters. The model described here holds true for the vast majority of PGCs; in healthy wild-type clutches, about 1% of cells is found in ectopic anterior positions and, in some cases, we see posterior trailing cells that do not align to the pronephros, but are found in the segmental plate.

the dorsoventral axis (reviewed in Mullins, 1999; Schier and average 27% of the cells are found posterior of the main cluster Talbot, 1998; Solnica-Krezel, 1999). (Table 1) and the majority of these ectopic cells is found in the Loss of function of the zebrafish chordin homolog chordino, tail around the blood-forming region (not shown). At the 60% a BMP antagonist, leads to a reduction in dorsal and anterior and 80% epiboly stages, the distribution of PGCs relative to tissues and concomitant expansion of the ventral-posterior the dorsal side of the embryo is normal (Table 1) reflecting domain (Hammerschmidt et al., 1996; Schulte-Merker et al., more or less normal execution of migration steps I to III. 1997). In 24-hour-old chordino mutant embryos, we find PGCs However, in 2- and 8-somite-stage chordino embryos, we find in ectopic locations (Fig. 4A, B). While many of the cells arrive more PGCs in the expanded ventral-posterior region of the at the correct position around the level of the 8th somite, on embryo (Table 1; Fig. 4C,D). Thus, in chordino mutants, stage Primordial germ cell migration in zebrafish 5301

Fig. 4. Migration of PGCs in ventralized chordino mutant embryos. The PGCs are labeled using the vasa probe (dark blue) and other structures are labeled in red with the probes indicated. (A,B) Ectopic PGCs are found around the expanded blood forming region in the tail of chordino mutants at 24 hpf; (A) lateral view, (B) dorsal view. (C,D) At the 8- somite stage (stained with myoD and gata2 labeling the Intermediate Cell Mass), more PGCs are found in the expanded ventral-posterior positions in chordino mutants (D) than in wild-type siblings (C) (see Fig. 5. The PGCs are found in random dorsoventral positions in also Table 1). dorsalized swirl mutants, but all are located at the same anteroposterior level. All embryos were stained with vasa in blue and V of PGC migration (migration towards the anterior) is a second staining in red was done with the probes indicated. (A,B) At specifically affected. It is possible that the expansion of the 1-somite stage, an alignment of PGCs anterior to trunk mesoderm (labeled by papc and myoD) is observed in lateral view of wild-type ventral-posterior fates disrupts the formation of graded (A) and swirl mutant (B) embryos (anterior up, dorsal right), but information that normally directs trailing PGCs towards the PGCs are found all around the dorsoventral axis in swirl mutant anterior. Additionally, it could be that the deficiency in anterior embryos. (C,D) Lateral view of 5-somite-stage embryos stained with structures leads to a decrease of a signal that normally attracts pax2.1 and myoD (here only the adaxial cells are labeled by this PGCs. Hence, instead of migrating anteriorly the cells remain probe). PGCs are still dispersed around the circumference of the in a ventral-posterior position and are finally located in ectopic embryo in swirl mutants (D); they do not cluster together as they do positions in the tail. in wild-type embryos at this stage (C) (see Fig. 1I for a dorsal view The swirl mutation, which reveals the function of the of wild-type). (E,F) PGCs have converged towards the dorsal in wild- zebrafish BMP2b gene, affects dorsoventral patterning of the type (E), but not swirl mutant (F) embryos at 80% epiboly. Embryos embryo in an opposite manner (Kishimoto et al., 1997; Mullins are shown in vegetal view with the dorsal marked by chordino expression in the wild-type (E) and circumferential chordino et al., 1996). swirl mutant embryos display a strong expression in the swirl mutant (F). dorsalization phenotype leading to formation of somites around the circumference of the embryo. These mutants show a striking PGC phenotype at the 1- and 5-somite stages with leading to a decrease in directional movement of the PGCs the cells located in a line along the expanded paraxial towards the dorsal midline of the embryo (Fig. 3A, step I). The mesoderm all around the embryo (Fig. 5A-D). This reduction convergence movement defects shared with somatic cells of convergence of PGCs towards the dorsal midline can be position the PGCs in a random dorsoventral position. Second, traced back to mid-gastrulation stages; more cells are found in ventral mesodermal fates are not specified in swirl mutants lateral and ventral positions at the 80% epiboly stage and the (e.g., expression of the pronephric markers pax 8 and pax2.1 initial ‘square’ shape of PGC clusters is maintained in about is not detected (data not shown and Mullins et al., 1996)) so half of the mutant embryos but is lost in all wild-type siblings the lateral edge of the mesoderm along which ventral PGCs are by this stage (Fig. 5E,F). Remarkably, while the PGCs appear normally found (e.g., Fig. 1B, lower panel) is not formed, to be located in random positions relative to the dorsoventral eliminating the option of PGCs to align parallel to the axis at the 1- and 5-somite stages, they are found at the correct anteroposterior axis. In contrast, in swirl mutants, the anterior anteroposterior level of the first somite (Fig. 5A,B). Thus, the border of the trunk paraxial mesoderm is formed and extends PGCs can properly respond to anteroposterior information, ventrally all around the embryo. Thus, PGCs at any position which extends around the entire embryo in swirl mutants. around the mutant embryo are close to this border as it forms We interpret the swirl phenotype as a manifestation of during gastrulation so that the alignment along the anterior defects in several processes required for normal PGC border of the trunk mesoderm (step IIIa, Fig. 3C) takes place migration. First, swirl mutants show reduced convergence around the circumference of the embryo. movements (Mullins et al., 1996; Solnica-Krezel, 1999) Importantly, the lack of ventral fates in swirl mutants affects 5302 G. Weidinger and others another movement normally found in wild-type embryos. Between the 1- and 5-somite stages, anteriorly located PGCs normally migrate laterally to form a cluster on either side of the axis (Fig. 3D, step IV). We do not observe this movement in swirl embryos where the random distribution of PGCs around the circumference of the embryo does not change from the 1- to the 5-somite stage. A likely explanation for this is that a lateral attracting target tissue located at the anteroposterior level of the first somite is not specified in swirl mutants, leaving the PGCs with no cues required for migration step IV. Defining intermediate targets for PGCs Ð migration along the lateral border of the mesoderm An intermediate target at the lateral border of the mesoderm that the PGCs could align to during migration step IIIb (Fig. 1A-C, bottom panels) is the Intermediate Cell Mass (ICM) which gives rise to blood and blood vessels and expresses the gata2 marker (Detrich et al., 1995). However, at the 7-somite stage, gata2 is expressed at a distance internally to the PGC line and the migration of PGCs in the cloche mutant that affects hematopoiesis from early stages on (Stainier et al., 1995) is not affected (data not shown). These findings suggest Fig. 6. PGC migration phenotype of that the ICM is not required for lateral alignment of PGCs. knypek and floating head mutant The possibility that the posterior PGCs align to the lateral embryos. All embryos were stained border of the endoderm was ruled out by analyzing the with vasa in dark blue and other expression pattern of the endodermal marker fork head-2 probes in red or light blue as (Odenthal and Nüsslein-Volhard, 1998) relative to the indicated. (A) Trailing PGCs align migrating cells (data not shown). along the abnormally located Another option for an intermediate target along which pronephros (stained with pax2.1, adaxial cells and somites with myoD) posterior PGCs align could be the developing pronephric in knypek mutant embryos at the 7- system or simply the lateral edge of the mesoderm. Several somite stage. (B,C) Normal arrangement of PGCs in wild-type (B) observations are consistent with a role for these structures in and floating head mutant (C) embryos at the 8-somite stage organizing the posterior PGCs. First, from the 1-somite stage stained with myoD and pax2.1. (D,E) Dorsal view of embryos at on posterior PGCs are found aligning at the lateral aspect of the 90% epiboly stage showing that dorsally located PGCs do not the mesoderm (Fig. 1F,I). When the lateral border of the migrate away from the dorsal midline in floating head mutant mesoderm and the pronephros are missing, as is the case in embryos. (D) In wild-type embryos, PGCs are not found at the swirl mutant embryos, the lateral alignment of trailing PGCs region of the forming notochord marked by lack of papc staining, is absent. In addition, alterations in the position of the border but they extend into the dorsal midline that expresses papc in of the mesoderm relative to the dorsal midline can be followed floating head mutant embryos (E). by corresponding alterations in the position of the PGCs. For example, in knypek mutant embryos, which are defective in mesoderm (Fig. 3C, step IIIa, Fig. 6D), but are absent from the convergence and extension movements (Solnica-Krezel et al., dorsal midline (Fig. 3B, step II, Fig. 6D). 1996), the pronephros is found far more lateral than in wild- The exclusion of PGCs from the midline suggests a role for type embryos, but still the posterior PGCs properly align to it the notochord in repelling the PGCs. To test this, we examined (Fig. 6A). PGC migration in floating head mutants in which the Currently, no mutation is known that specifically affects the notochord is replaced by paraxial mesodermal fates (Halpern development of the pronephric anlage as early as the 80% et al., 1995; Talbot et al., 1995). Clearly evident at 60% epiboly epiboly to 1-somite stages. In the no isthmus mutant, which and shown in Fig. 6E for the 90% epiboly stage, PGCs can be shows a late phenotype in the pronephric system (Brand et al., observed in an ectopic position at the dorsoventral level of the 1996), the early alignment of the ventral PGCs to the notochord in floating head mutants (3.1±2.3 cells (n=10) pronephric anlage is not affected (data not shown). Therefore, compared with 0.4±0.8 in wild-type siblings (n=22)). while the developing pronephros is a good candidate for an Therefore, the notochord appears to be required directly or intermediate PGC target, we could not prove this point using indirectly for exclusion of PGCs from the midline during early mutations specifically affecting this structure. gastrulation (Fig. 3B, step II). This phenotype is however only transiently observed; from about the 2-somite stage on when Defining intermediate targets for PGCs Ð alignment the PGCs start to migrate laterally no ectopic cells are found along the anterior border of the trunk paraxial at the midline of floating head mutants. In addition, both the mesoderm anterior alignment step (Fig. 3C, step IIIa) and the formation As described above, PGCs that at the beginning of gastrulation of the lateral clusters around the 5-somite stage (Fig. 3D, step are positioned closer to the dorsal side align at the first somite IV) are normal (data not shown and Fig. 6B,C). We therefore level along the border between the trunk and head paraxial conclude that differentiated axial mesoderm is not required for Primordial germ cell migration in zebrafish 5303

Fig. 7. Ectopic PGCs are found in spadetail mutant embryos. All embryos were stained with vasa in blue and other probes in red as indicated. (A,B) Ectopic anterior PGCs are seen between the midbrain-hindbrain boundary and the otic vesicle in 24 hpf stage spadetail mutant embryos and ectopic posterior PGCs are frequently observed in the tail; (A) lateral view, (B) dorsal view. (C,D) PGC alignment at the anterior border of the papc expresssion domain at the 80% epiboly stage as seen in a dorsal view of a wild- type embryo (C) is lost in spadetail mutants (D). (E,F) At the 3- somite stage, alignment of PGCs at Fig. 8. PGCs maintain their fate at ectopic locations. All embryos st the level of the 1 were stained with the vasa probe in blue. (A, B) Double staining of somite as seen in wild- wild-type (A) and spadetail mutant (B) embryos with vasa and type embryos (E) is hoxa2 both in blue shows that anterior ectopic PGCs (arrows) are lost in spadetail located at the anteroposterior level of the 2nd branchial arch mutants (F) stained (brackets) at 24 hpf. (C, D) Ectopic vasa-expressing cells maintain with myoD, papc and PGC morphology. Cross-section of a spadetail mutant (C) and a pax8. (G,H) At the 6- wild-type (D) embryo at the indicated levels. Note that both the somite stage (embryos ectopic vasa-positive cells in C and the PGCs located in the correct stained with myoD and position in D are large and show a distinct nuclear shape (see insert). pax2.1), the main clusters of PGCs located at the 1st to the we followed the migration of the PGCs in spadetail mutants. 3rd somite in wild-type The spadetail gene encodes a T-box protein important for embryos (G) are found proper development of trunk paraxial mesoderm (Griffin et al., closer to the otic 1998). In spadetail mutants, the expression of genes that mark placodes in spadetail the anterior border of the trunk paraxial mesoderm, like papc mutants (H) and ectopic anterior PGCs and paraxis, is already severely reduced at early stages of are found mainly in gastrulation (Ho and Kane, 1990; Shanmugalingam and between the midbrain- Wilson, 1998; Yamamoto et al., 1998). In 24 hpf spadetail hindbrain boundary mutants, we observe PGCs in several ectopic positions (Fig. and the otic placodes in 7A,B). In 72% of the spadetail mutants, PGCs are located in the mutants. ectopic posterior positions along the yolk extension or in the mutant tail (on average 9% of all PGCs show this behavior, Table 2, Fig. 7A). Strikingly, in 96% of the mutant embryos, defining the anteroposterior level of alignment nor for repelling we find PGCs located in an ectopic anterior position at 24 hpf PGCs away from the midline at the 1- to 5-somite stage. (on average 16% of all PGCs are located anteriorly; Fig. 7A,B; Thus, an attractive possibility is that cues originating in the Table 2). Virtually all of these ectopic anterior PGCs are found paraxial mesoderm are responsible for the early alignment of at the same anteroposterior position at the level of the 2nd PGCs at the border of the trunk mesoderm. To test this idea, branchial arch (Fig. 8A,B). The appearance of ectopic anterior 5304 G. Weidinger and others

PGCs can be traced back to loss of PGC alignment at the head- To verify that the ectopic cells, in addition to expressing trunk mesoderm border: at the 80% epiboly stage, dorsal PGCs vasa, preserve other PGC characters, we examined their extend towards the anterior in most of the mutant embryos (Fig. morphology. As can be seen in Fig. 8C,D, the morphology of 7C,D). Consequently, ectopic anterior PGCs can be seen in the ectopic vasa-expressing cells in spadetail mutants is similar spadetail mutant embryos at the 3- and 6-somite stages (Fig. to that of PGCs in normal positions. Both the ectopic PGCs 7E-H; Table 2). At the 3-somite stage, the normal alignment and cells in the normal position are very large and, in most of PGCs at the first somite level is not observed. Instead, all cases, show a distinct lobular nuclear shape (Fig. 8C, insert). PGCs are found at the lateral border of the mesoderm These cellular features are characteristic of PGCs in extending into ectopic anterior regions (Fig. 7F). Thus, in (e.g. in B. conchonius (Gevers et al., 1992)). Thus, the ectopic spadetail mutant embryos, step IIIa of PGC migration is cells maintain vasa gene expression and exhibit morphological affected. characteristics typical for PGCs, suggesting preservation of An interesting open question is why all of the ectopic their fate in the ectopic location at least for the first 48 hours anterior cells in spadetail mutants end up at a defined position of development. We believe that at later stages of development at the anteroposterior level of the 2nd branchial arch. Two the ectopic PGCs die and consider it less likely that at those points relevant to this phenomenon can be made. First, the late stages they differentiate and join somatic tissues. However, definition of this position of ectopic PGC accumulation is not since we did not follow the final fate of the cells, the possibility a result of spadetail loss of function. In wild-type embryos, we of very late differentiation of ectopic PGC was not ruled out. also observed rare cases of ectopic anterior cells (e.g. Fig. 1I, arrow; 1.3% of PGCs in wild type versus 16% in spadetail mutants). The position of these ectopic cells is identical to that DISCUSSION seen in the mutants. Second, tissue sections containing ectopic PGCs in this region in spadetail mutants reveal close cellular In this work, we followed the migration of the primordial germ interaction between the PGCs and surrounding branchial arch cells in early wild-type embryos and characterized six discrete tissue (Fig. 8C). Thus, the PGCs may share cell adhesion steps of cell movements that bring the PGCs to their position properties with the somatic cells located at this position at 24 hours of development. Some of these movements are resulting in a defined accumulation point for anterior PGCs. obviously shared with somatic cells (e.g., step I, convergence towards the dorsal). For other steps, for example step IV The determination state of early migrating PGCs in (formation of two lateral PGC clusters) no corresponding zebrafish movements of somatic cells have been described, suggesting The migrating vasa-expressing cells face conflicting that these movements reflect active PGC migration relative to requirements related to their response for cues provided by the the surrounding tissue. Analysis of the behavior of PGCs in environment. On one hand, as shown above, the PGCs respond mutant embryos allows us to suggest which tissues are required to signals provided by somatic tissues guiding them towards for proper PGC migration during early stages of the gonad. On the other hand, during their migration, the PGCs embryogenesis and to address questions related to PGC fate. have to ignore signals that would lead to their differentiation into other cell types and to loss of totipotency. Migration of PGCs Ð solutions for fish-specific The abnormal migration of PGCs in spadetail mutant problems embryos allows us to address the question of the determination When compared with other model organisms, the zebrafish state of the early migrating PGCs in fish. In these mutants, shows a unique process of PGC migration. While in mouse, ectopic cells can be observed starting from the 80% epiboly Drosophila, Xenopus and urodeles, the migration starts from stage until at least 2 days of development (data not shown). the same position in each embryo with the PGCs being directed Since the number of ectopic vasa-positive cells does not towards the target in a predictable path, the situation in decrease between the 24 hpf stage and the 48 hpf stage (Table zebrafish appears more complex. Based on our observations we 2), it appears that vasa-expressing cells can survive in ectopic suggest a model that helps understanding how PGCs that can locations for at least 2 days. originate at any position relative to the dorsal tissue are able to One possible explanation for this phenomenon is that the arrive at a specific dorsoventral and anteroposterior level within PGC fate is lost in the ectopic location and that the observed the first 12 hours of development. vasa signal reflects an exceptional stability of this mRNA. To According to our model, the PGCs, which by the beginning test this, we injected wild-type embryos with an in vitro of gastrulation are positioned close to the margin of the transcribed vasa mRNA. Early overexpression of vasa mRNA blastoderm, initially follow the general behavior described for did not lead to a change in the number of vasa-expressing cells hypoblast cells, which undergo dorsal convergence. Starting at observed in 24 hpf embryos. Importantly, the injected vasa early gastrulation, the PGCs align to structures that define the mRNA is not exceptionally stable, so that 24 hours border of the trunk mesoderm (step IIIa and IIIb). This simple postinjection no traces of the injected mRNA can be detected requirement puts all the PGCs in proximity to an intermediate (data not shown). The maintenance of the vasa signal by the target irrespective of their initial position (Fig. 9A). Through ectopic cells must therefore result from preservation of PGC- such a mechanism an ordered distribution of PGCs is achieved specific qualities, which might either preferentially protect the relatively early in gastrulation in contrast with the apparent vasa mRNA from degradation (as described in C. elegans for random positioning of PGCs a few hours earlier. This early germ-cell-specific RNAs (Seydoux and Fire, 1994)) or, more arrangement is apparent from the 80% epiboly stage until the likely, could keep on synthesizing vasa mRNA during 1-somite stage when the ventrally located cells are aligned embryogenesis. along the pronephros and cells that originated in the dorsal Primordial germ cell migration in zebrafish 5305

60% epiboly align at the level of the first somite (Fig. 9B). The importance A of the early alignment is most strikingly demonstrated in spadetail mutants, which as judged by molecular markers, lack a defined anterior border of the trunk paraxial mesoderm; here the dorsally located PGCs do not align at the trunk-head mesoderm border and are therefore found in ectopic anterior locations. Specific early elimination of the pronephros is not ventral dorsal ventral possible using any existing mutations, so we were not able to directly test a possible interaction of PGCs with this tissue. B bud However, since PGCs are found in close proximity of the pronephric anlage from early stages onwards, and since other ventral tissues like the intermediate cell mass apparently do not play a role in directing PGC migration, the pronephros remains an attractive candidate to serve as an intermediate target for PGCs. Interestingly, in urodeles, positioning of PGCs in the lateral plate mesoderm has been described. While PGCs in urodeles arise in ventral positions (i.e. not random relative to the dorsal), during somitogenesis these cells are found in a position reminiscent of that taken by posterior zebrafish PGCs which align along the pronephros (Nieuwkoop and Sutasurya, 1979). C 3 somites Beginning after the 1-somite stage, we observe directed migration of cells towards a lateral tissue at the anteroposterior level of the first somite. Both the posteriorly located trailing cells as well as the anteriorly aligned cells can be observed to migrate towards this position (Fig. 9C). The existence of an attractive signaling center at the lateral anterior part of the trunk mesoderm is consistent with the normal lateral migration of PGCs (step IV) in floating head embryos, which argues against repulsion by the notochord at this stage. The swirl and the spadetail phenotypes also argue for the existence of such a signaling center. The absence of the laterally directed movements of the PGCs in swirl mutants could reflect loss of ventral and lateral cell fates including that of the putative Fig. 9. A model for a transition from a random distribution of signaling center. In contrast, in spadetail mutants, this PGCs prior to gastrulation to an organized arrangement of attraction center is not lost, since at 24 hpf the primordial germ migrating cells. Schematic drawings show the tissues that provide cells are not randomly distributed anterior to the normal cluster positional cues for the early migrating PGCs with the notochord in as seen at the 1-somite stage. Instead, in spadetail embryos, gray, trunk mesoderm in light blue and the PGCs in red. Based on most of the PGCs appear to migrate back towards the main analysis of multiple embryos, the PGCs are drawn in different cluster while some migrate towards a specific position at the possible positions; in any single embryo, PGCs are found in only level of the 2nd branchial arch. some of these locations. Before gastrulation, the PGC clusters are found in random dorsoventral positions close to the blastoderm Specification and maintenance of PGC fate margin. (A) At early gastrulation stages, PGCs are found scattered around the border of the trunk mesoderm, which they meet or The identification of the zebrafish vasa homolog and its early migrate towards (indicated by black arrows), gradually forming a localization to the PGCs (Yoon et al., 1997) suggest that clear alignment pattern by the end of gastrulation. Cells located specification of the PGCs in zebrafish occurs by a mechanism directly at or close to the dorsal midline where the notochord is of asymmetric localization of cytoplasmic factors including forming are excluded from this region and occupy a more lateral mRNAs. Interestingly, when vasa mRNA was injected into position. This tendency to align at the borders of the trunk early embryos, we did not detect a change in the number of mesoderm leads to an organized arrangement of all PGCs relative PGCs. We conclude that vasa mRNA per se is not a limiting to early developing structures during gastrulation. (B) At the tail factor for PGC specification. It is possible that, in addition to bud stage, PGCs are located in a line along the anterior border of vasa, other cytoplasmic components need to be localized in the the trunk mesoderm or align along the pronephros at the lateral border of the mesoderm. These posteriorly located cells migrate future PGCs or that an assembly of an active complex of towards the anterior. (C) We propose that, at the 2- to 3-somite molecules takes place during oogenesis. Indications for the stage, a signaling center (dark blue) is established that attracts temporal requirement for vasa activity come from Drosophila PGCs located in its vicinity. This leads to migration of PGCs where the Vasa protein is localized to the posterior pole and is located at the anterior trunk mesoderm border towards more lateral functionally required during oogenesis for patterning of the positions and may participate in directing posterior trailing cells to abdomen (Hay et al., 1988; Lasko and Ashburner, 1988; the anterior. Lehmann and Nüsslein-Volhard, 1991). Once specified, the PGCs should maintain their totipotency and ignore differentiation signals other than signals important 5306 G. Weidinger and others for their differentiation as gametes. The issue of PGC fate has REFERENCES been investigated extensively in Drosophila. By following the fate of labeled transplanted PGCs, it was found that a large Abdelilah, S., Solnica-Krezel, L., Stainier, D. and Driever, W. (1994). proportion (up to 50%) of the PGCs end up in regions outside Implications for dorsoventral axis determination from the zebrafish mutation janus. Nature 370, 468-471. of the gonad. These cells, however, never differentiated as Anderson, R., Fassler, R., Georges-Labouesse, E., Hynes, R. O., Bader, B. somatic cells, were never incorporated into adjacent tissues L., Kreidberg, J. A., Schaible, K., Heasman, J. and Wylie, C. (1999). and eventually died (Kobayashi et al., 1993; Technau and Mouse primordial germ cells lacking β1 integrins enter the germline but fail Campos-Ortega, 1986; Underwood et al., 1980). In the to migrate normally to the gonads. Development 126, 1655-1664. Bernex, F., de Sepulveda, P., Kress, C., Elbaz, C., Delouis, C. and Panthier, experiments described here, we did not label the PGCs with a J. (1996). Spatial and temporal patterns of c-kit-expressing cells in WlacZ/+ stable lineage marker, but rather relied on vasa gene and WlacZ/WlacZ mouse embryos. Development 122, 3023-3033. expression. Our ability to follow cells located in several Brand, M., Heisenberg, C. P., Jiang, Y. J., Beuchle, D., Lun, K., Furutani- different ectopic locations for a few days (e.g., in chordino and Seiki, M., Granato, M., Haffter, P., Hammerschmidt, M., Kane, D. A. in spadetail mutants) coupled with the fact that the total et al. (1996). Mutations in zebrafish genes affecting the formation of the boundary between midbrain and hindbrain. Development 123, 179-190. number of vasa-positive cells is similar in wild-type and Detrich, H. W., 3rd, Kieran, M. W., Chan, F. Y., Barone, L. M., Yee, K., mutant embryos argues for the reliability of vasa as a stable Rundstadler, J. A., Pratt, S., Ransom, D. and Zon, L. I. (1995). marker for PGCs. At least during the first 48 hours of Intraembryonic hematopoietic cell migration during development, the ectopic cells maintain some PGC characters development. Proc. Natl. Acad. Sci. USA 92, 10713-10717. Gevers, P., Dulos, J., Schipper, H. and Timmermans, L. (1992). Origin of as determined by their morphology and by the fact that they primordial germ cells, as characterized by the presence of nuage, in embryos express the vasa gene. Our findings are therefore consistent of the fish Barbus conchonius. Eur. J. Morphol. 30, 195-204. with those from Drosophila suggesting that PGCs normally Godin, I., Wylie, C. and Heasman, J. (1990). Genital ridges exert long-range do not contribute to other cell lineages. effects on mouse primordial germ cell numbers and direction of migration These results differ from results in mouse and Xenopus in culture. Development 108, 357-363. Griffin, K. J., Amacher, S. L., Kimmel, C. B. and Kimelman, D. (1998). (Stevens, 1970; Wylie et al., 1985). In these systems, Molecular identification of spadetail: regulation of zebrafish trunk and tail transplantation of migrating germ cells into ectopic locations mesoderm formation by T-box genes. Development 125, 3379-3388. led to their differentiation into a variety of somatic tissues. A Halpern, M., Thisse, C., Ho, R., Thisse, B., Riggleman, B., Trevarrow, B., possible explanation for these conflicting results may be that Weinberg, E., Postlethwait, J. and Kimmel, C. (1995). Cell-autonomous shift from axial to paraxial mesodermal development in zebrafish floating the behavior of the cells was studied at different stages of their head mutants. Development 121, 4257-4264. development. We investigated the determination state of the Hammerschmidt, M., Pelegri, F., Mullins, M. C., Kane, D. A., van Eeden, PGCs at an early stage: the abnormally migrating cells that we F. J., Granato, M., Brand, M., Furutani-Seiki, M., Haffter, P., followed in mutant embryos get into ectopic positions starting Heisenberg, C. P. et al. (1996). dino and mercedes, two genes regulating as early as mid-gastrulation and thus have not been subjected dorsal development in the zebrafish embryo. Development 123, 95-102. Hauptmann, G. (1999). Two-color detection of mRNA transcript localizations to the normal signaling that would be provided by the gonad in fish and fly embryos using alkaline phosphatase and β-galactosidase or tissues adjacent to it. The transplanted PGCs in the mouse conjugated . Dev. Gen. Evol. 209, 317-321. and in Xenopus have been obtained from stages of migration Hauptmann, G. and Gerster, T. (1994). Two-color whole-mount in situ that most likely correspond to later steps in zebrafish. Since the hybridization to vertebrate and Drosophila embryos. Trends Genet. 10, 266. Hay, B., Jan, L. and Jan, Y. (1988). A protein component of Drosophila polar PGCs have to respond to signals from the somatic tissue of the granules is encoded by vasa and has extensive sequence similarity to ATP- gonad (e.g. Mukai et al., 1999), their resistance to dependent helicases. Cell 55, 577-587. differentiation cues should be relieved at later stages. We Helde, K. A., Wilson, E. T., Cretekos, C. J. and Grunwald, D. J. (1994). therefore propose that these seemingly conflicting results Contribution of early cells to the fate map of the zebrafish gastrula. Science simply reflect two different stages in the differentiation of the 265, 517-520. Ho, R. K. and Kane, D. A. (1990). Cell-autonomous action of zebrafish spt- early PGCs. 1 mutation in specific mesodermal precursors. Nature 348, 728-730. Howard, K. (1998). Organogenesis: Drosophila goes gonadal. Curr. Biol. 8, Conclusions R415-417. In this work, we provide a detailed description of germ cell Jaglarz, M. and Howard, K. (1994). Primordial germ cell migration in is controlled by somatic tissue. Development 120, migration in zebrafish and propose a model to explain the 83-89. observed cell behavior in wild-type and mutant embryos. Jowett, T. and Lettice, L. (1994). Whole-mount in situ hybridizations on Future studies aimed at testing and refining this first model will zebrafish embryos using a mixture of digoxigenin- and fluorescein-labelled utilize newly identified mutations specifically affecting the probes. Trends Genet. 10, 73-74. process of germ cell migration and will be enhanced by Kimmel, C. B., Kane, D. A., Walker, C., Warga, R. M. and Rothman, M. B. (1989). A mutation that changes cell movement and cell fate in the generation of fish lines that allow the migration process to be zebrafish embryo. Nature 337, 358-362. followed in live embryos. Kishimoto, Y., Lee, K., Zon, L., Hammerschmidt, M. and Schulte-Merker, S. (1997). The molecular nature of zebrafish swirl: BMP2 function is We thank colleagues in the zebrafish field for probes and the labs essential during early dorsoventral patterning. Development 124, 4457- of Christiane Nüsslein-Volhard, Matthias Hammerschmidt and 4466. Wolfgang Driever for zebrafish strains. We thank Wolfgang Driever Kobayashi, S., Kitamura, T., Sasaki, H. and Okada, M. (1993). Two types for help and support, Gerlinda Wussler for taking care of the fish of pole cells are present in the Drosophila embryo, one with and one facility and other members of the department for helpful discussions. without splicing activity for the third P-element intron. Development 117, 885-893. We are grateful to Randy Cassada, Iain Drummond and Lila Solnica- Krauss, S., Johansen, T., Korzh, V. and Fjose, A. (1991). Expression pattern Krezel for critical reading of the manuscript. This work was supported of zebrafish pax genes suggests a role in early brain regionalization. Nature by grants from the European Commission TMR program 353, 267-270. (ERBFMBICT983315), from the Deutsche Forschungsgemeinschaft Lasko, P. and Ashburner, M. (1988). The product of the Drosophila gene (RA863/1-1) and from the Landesschwerpunkt Baden-Württemberg. vasa is very similar to eukaryotic initiation factor-4A. Nature 335, 611-617. Primordial germ cell migration in zebrafish 5307

Lehmann, R. and Nüsslein-Volhard, C. (1991). The maternal gene nanos has Solnica-Krezel, L. (1999). Pattern formation in zebrafish-fruitful liaisons a central role in posterior pattern formation of the Drosophila embryo. between embryology and genetics. Curr. Top. Dev. Biol. 41, 1-35. Development 112, 679-691. Solnica-Krezel, L., Stemple, D. L. and Driever, W. (1995). Transparent Matsui, Y., Zsebo, K. and Hogan, B. (1990). Embryonic expression of a things: cell fates and cell movements during early embryogenesis of haematopoietic growth factor encoded by the Sl locus and the ligand for c- zebrafish. BioEssays 17, 931-939. kit. Nature 347, 667-669. Solnica-Krezel, L., Stemple, D. L., Mountcastle-Shah, E., Rangini, Z., Miller-Bertoglio, V. E., Fisher, S., Sanchez, A., Mullins, M. C. and Neuhauss, S. C., Malicki, J., Schier, A. F., Stainier, D. Y., Zwartkruis, Halpern, M. E. (1997). Differential regulation of chordin expression F., Abdelilah, S. et al. (1996). Mutations affecting cell fates and cellular domains in mutant zebrafish. Dev. Biol. 192, 537-550. rearrangements during gastrulation in zebrafish. Development 123, 67-80. Moore, L. A., Brohier, H. T., Dore, M. V., Lunsford, L. B. and Lehmann, Stainier, D. Y., Weinstein, B. M., Detrich, H. W. 3rd., Zon, L. I. and R. (1998). Identification of genes controlling germ cell migration and Fishman, M. C. (1995). Cloche, an early acting zebrafish gene, is required embryonic gonad formation in Drosophila. Development 125, 667-678. by both the endothelial and hematopoietic lineages. Development 121, 3141- Mukai, M., Kashikawa, M. and Kobayashi, S. (1999). Induction of 3150. indora expression in pole cells by the mesoderm is required for female Stevens, L. (1970). The development of transplantable teratocarcinomas from germ-line development in Drosophila melanogaster. Development 126, intratesticular grafts of pre-and post-implantation embryos. Dev. Biol. 21, 1023-1029. 364-382. Mullins, M. C. (1999). Embryonic axis formation in the zebrafish. Methods Talbot, W. S., Trevarrow, B., Halpern, M., Melby, A. E., Farr, G., Cell Biol. 59, 159-178. Postlethwait, J. H., Jowett, T., Kimmel, C. B. and Kimelman, D. (1995). Mullins, M. C., Hammerschmidt, M., Kane, D. A., Odenthal, J., Brand, A homeobox gene essential for zebrafish notochord development. Nature M., van Eeden, F. J., Furutani-Seiki, M., Granato, M., Haffter, P., 378, 150-157. Heisenberg, C. P. et al. (1996). Genes establishing dorsoventral pattern Technau, G. and Campos-Ortega, J. (1986). Lineage analysis of transplanted formation in the zebrafish embryo: the ventral specifying genes. individual cells in embryos of Drosophila melanogaster III. Commitment Development 123, 81-93. and proliferation capabilities of pole cells and midgut progenitors. Roux's Nieuwkoop, P. D. and Sutasurya, L. A. (1979). Primordial germ cells in the Arch. Dev. Biol. 195, 489-498. chordates. Embryogenesis and phylogenesis. pp. 87-92. Cambridge: Underwood, E. M., Caulton, J. H., Allis, C. D. and Mahowald, A. P. (1980). Cambridge University Press. Developmental fate of pole cells in Drosophila melanogaster. Dev. Biol. 77, Odenthal, J., Haffter, P., Vogelsang, E., Brand, M., van Eeden, F. J., 303-314. Furutani-Seiki, M., Granato, M., Hammerschmidt, M., Heisenberg, C. van Doren, M., Brohier, H. T., Moore, L. A. and Lehmann, R. (1998). P., Jiang, Y. J. et al. (1996). Mutations affecting the formation of the HMG-CoA reductase guides migrating primordial germ cells. Nature 396, notochord in the zebrafish, Danio rerio. Development 123, 103-115. 466-469. Odenthal, J. and Nüsslein-Volhard, C. (1998). fork head domain genes in Warga, R. and Nüsslein-Volhard, C. (1998). spadetail-dependent cell zebrafish. Dev Genes Evol 208, 245-258. compaction of the dorsal zebrafish blastula. Dev. Biol. 203, 116-121. Olsen, L. C., Aasland, R. and Fjose, A. (1997). A vasa-like gene in zebrafish Weinberg, E. S., Allende, M. L., Kelly, C. S., Abdelhamid, A., Murakami, identifies putative primordial germ cells. Mech. Dev. 66, 95-105. T., Andermann, P., Doerre, O. G., Grunwald, D. J. and Riggleman, B. Pfeffer, P. L., Gerster, T., Lun, K., Brand, M. and Busslinger, M. (1998). (1996). Developmental regulation of zebrafish MyoD in wild-type, no tail Characterization of three novel members of the zebrafish Pax2/5/8 family: and spadetail embryos. Development 122, 271-280. dependency of Pax5 and Pax8 expression on the Pax2.1 (noi) function. Westerfield, M. (1995). The Zebrafish Book. Oregon: University of Oregon Development 125, 3063-3074. Press. Prince, V. E., Joly, L., Ekker, M. and Ho, R. K. (1998). Zebrafish hox genes: Williamson, A. and Lehmann, R. (1996). Germ cell development in genomic organization and modified colinear expression patterns in the trunk. Drosophila. Ann. Rev. Cell Dev. Biol. 12, 365-391. Development 125, 407-420. Wylie, C. (1999). Germ cells. Cell 96, 165-174. Rongo, C., Broihier, H., L, M., Van Doren, M., Forbes, A. and Lehmann, Wylie, C., Heasman, J., Snape, A., O'Driscoll, M. and Holwill, S. (1985). R. (1997). Germ plasm assembly and germ cell migration in Drosophila. Primordial germ cells of Xenopus laevis are not irreversibly determined Cold Spring Harb. Symp. Quant. Biol. 62, 1-11. early in development. Dev. Biol. 112, 66-72. Schier, A. and Talbot, W. (1998). The zebrafish organizer. Curr. Opin. Genet. Yamamoto, A., Amacher, S. L., Kim, S. H., Geissert, D., Kimmel, C. B. Dev. 8, 464-471. and De Robertis, E. M. (1998). Zebrafish paraxial protocadherin is a Schulte-Merker, S., Lee, K. J., McMahon, A. P. and Hammerschmidt, M. downstream target of spadetail involved in morphogenesis of gastrula (1997). The zebrafish organizer requires chordino. Nature 387, 862-863. mesoderm. Development 125, 3389-3397. Schüpbach, T. and Wieschaus, E. (1986). Maternal-effect mutations altering Yoon, C., Kawakami, K. and Hopkins, N. (1997). Zebrafish vasa homologue the anterior-posterior pattern of the Drosophila embryo. Roux's Arch. Dev. RNA is localized to the cleavage planes of 2- and 4-cell-stage embryos and Biol. 195, 302-317. is expressed in the primordial germ cells. Development 124, 3157-3165. Seydoux, G. and Fire, A. (1994). Soma-germline asymmetry in the Zhang, N., Zhang, J., Cheng, Y. and Howard, K. (1996). Identification and distributions of embryonic RNAs in . Development genetic analysis of wunen, a gene guiding Drosophila melanogaster germ 120, 2823-2834. cell migration. Genetics 143, 1231-1241. Shanmugalingam, S. and Wilson, S. W. (1998). Isolation, expression and Zhang, N., Zhang, J., Purcell, K., Cheng, Y. and Howard, K. (1997). The regulation of a zebrafish paraxis homologue. Mech. Dev. 78, 85-89. Drosophila protein Wunen repels migrating germ cells. Nature 385, 64-67.