ABSTRACT
During development of the vertebrate hindbrain regulatory gene expression is confined to precise segmental domains. Studies of cell lineage and gene expression suggest that establishment of these domains may involve a dynamic regulation of cell identity and restriction of cell movement between segments. We have taken a dominant negative approach to interfere with the function of Sek-1, a member of the Eph-related receptor tyrosine kinase family expressed in rhombomeres r3 and r5. In Xenopus and zebrafish embryos expressing truncated Sek-1, lacking kinase sequences, expression of r3/r5 markers occurs in adjacent even-numbered rhombomeres, in domains contiguous with r3 or r5. This disruption is rescued by fulllength Sek-1, indicating a requirement for the kinase domain in the segmental restriction of gene expression. These data suggest that Sek-1, perhaps with other Ephrelated receptors, is required for interactions that regulate the segmental identity or movement of cells.
INTRODUCTION
In many animal phyla, certain tissues are subdivided during embryogenesis into repeated morphological units, or segments, that then differentiate to generate a series of homologous structures. The molecular mechanisms underlying segmentation are best understood in the Drosophila embryo. In this system, the embryo is subdivided into presumptive segments at syncytial stages by a cascade of interactions between genes encoding transcription factors. This cascade establishes the expression of short-range signalling molecules in adjacent cells at the parasegment border, and these signals regulate the positional identity of neighbouring cells, as marked by segmental gene expression, and stabilise the boundary region (reviewed by Ingham and Martinez Arias, 1992; DiNardo et al., 1994). In contrast, little is known regarding the molecular mechanisms regulating the positional identity of cells during segmentation in the vertebrate embryo.
In recent years it has been shown that segmentation occurs in the vertebrate hindbrain and underlies the specification of nerves (Metcalfe et al., 1986; Hanneman et al., 1988; Lumsden and Keynes, 1989; Clarke and Lumsden, 1993) and of neural crest cells migrating to the branchial arches (Noden, 1983, 1988; Lumsden et al., 1991; Serbedzija et al., 1992; Sechrist et al., 1993). A molecular correlate of this cellular patterning is provided by the segment-restricted expression domains of genes, including Krox-20 and Hox genes, encoding transcription factors required for the formation or the anterior-posterior (A-P) specification of segments (reviewed by McGinnis and Krumlauf, 1992; Wilkinson, 1993). For example, the Krox-20 gene, required for r3 and r5 formation (Schneider-Maunoury et al., 1993), is up-regulated in the early neural plate prior to segmentation in two fuzzy domains that then become progressively sharper and restricted precisely to definitive r3 and r5 (Wilkinson et al., 1989; Nieto et al., 1991; Irving et al., 1995). This raises the question as to how the fuzzy expression domains are sharpened to become segment-restricted.
Studies in the chick embryo provide evidence for cellular mechanisms that might underlie the segmental restriction of gene expression. Transplantation experiments indicate that a regional specification of presumptive r4 occurs prior to segmentation (Guthrie et al., 1992). However, the clonal progeny of a single neuroepithelial cell labelled before segmentation disperse widely and frequently contribute to more than one segment (Fraser et al., 1990). After boundary formation most clonal progeny are confined to a single rhombomere (Fraser et al., 1990), but even at this stage some contribute to an adjacent segment (Birgbauer and Fraser, 1994). A potential mechanism for the restriction of cell movement is indicated by the finding that cells from r3 and r5 mix more readily with each other than they do with cells from r2, r4 and r6 (Guthrie et al., 1993). Similarly, cells from r2, r4 and r6 are more miscible with each other than they are with r3 or r5. Cell adhesion properties that alternate between rhombomeres may therefore restrict cell movement across boundaries, and cell sorting could contribute to the sharpening of gene expression domains. However, since individual cells are not irreversibly committed to a specific rhombomere there must also be a dynamic regulation of the segmental identity of cells.
Based upon the critical roles of receptor tyrosine kinases (RTKs) in transducing signals that regulate cell fate in other systems (reviewed by Pawson and Bernstein, 1990; Dickson and Hafen, 1994; van der Geer et al., 1994), genes mediating cell-cell interactions during hindbrain segmentation might be found among the RTK superfamily. Indeed, in recent work, we and others have identified a number of RTKs segmentally expressed in the hindbrain that are members of the Eph-related family. Eph-related genes make up the largest known family of RTKs and their expression in the developing and mature CNS suggests potential roles in neural development and function (for review see van der Geer et al., 1994). However, little is known regarding the roles of Eph-related RTKs. Recently, several ligands have been identified and shown to constitute a family of structurally related proteins, all anchored in the plasma membrane, either via a transmembrane domain or a glycosylphosphatidyl inositol linkage (Bartley et al., 1994; Beckmann et al., 1994; Cheng and Flanagan, 1994; Davis et al., 1994). It is therefore likely that Eph-related RTKs mediate cell contact-dependent signalling.
Studies of Eph-related RTKs in the mouse have shown that, in addition to expression elsewhere in the early embryo, Sek-1 (Nieto et al., 1992) and Sek-2/eck (Becker et al., 1994; Ganju et al., 1994; Ruiz and Robertson, 1994) are up-regulated in the hindbrain prior to segmentation, in pre-r3 plus pre-r5 and in pre-r4, respectively. In contrast, expression of Sek-3/nuk (Becker et al., 1994; Henkemeyer et al., 1994) and Sek-4 (Becker et al., 1994) is restricted to r3 plus r5 later, after segmentation. Three Eph-related zebrafish genes, rtk1-3 (Xu et al., 1994), also have segmental expression in the hindbrain (Q.X. and N.H., unpublished observations). These studies suggest that several members of the Eph family have segmentrestricted roles in transducing signals, possibly cell contactdependent, during hindbrain patterning.
To analyse the function of Sek-1 protein in the hindbrain we have taken a dominant negative approach. We find that microinjection of RNA encoding truncated Sek-1 receptor into Xenopus and zebrafish embryos leads to disruption of the spatial restriction of r3/r5 gene expression in the hindbrain. These data suggest that Sek-1, perhaps together with other Eph-related RTKs, mediates interactions that are required for the segmental restriction of gene expression, and we discuss possible roles of Sek-1 in this restriction.
MATERIALS AND METHODS
Cloning and sequencing of Sek-1 cDNAs
A neurula-stage Xenopus cDNA library was screened at high stringency (0.2× SSC, 60°C) with a probe including the kinase domain of mouse Sek-1. Positives were picked, subcloned into Bluescript, and sequenced using the dideoxy chain termination method, either with fragments generated by exonuclease deletion or using oligonucleotides corresponding to previously determined sequence.
Generation of RNA encoding full-length and mutant Sek-1
Sequences were subcloned into pSP64TK, a derivative of pSP64T (which flanks the coding region with β-globin untranslated sequences; Krieg and Melton, 1984) modified to include a Kozak consensus sequence at the initiation codon, which is within an NcoI site, and stop codons in all three frames immediately downstream from the coding sequence. The truncated mouse Sek-1 sequences correspond to a NcoI-BamHI fragment, and the truncated XSek-1 sequences correspond to an NcoI-DraI fragment. Both of these fragments start at the initiation codon and terminate just downstream from the transmembrane domain (as indicated in Fig. 1). Capped RNA was synthesised by the in vitro transcription of linearised plasmids (Moon and Christian, 1989), dissolved at 1 μg/μl and stored in aliquots at −70°C. The sizes of the encoded polypeptides were analysed by in vitro translation with rabbit reticulocyte lysate (Promega) and gel electrophoresis. Immediately before use for microinjection an aliquot of RNA was mixed with 10 μg/μl lysinated rhodamine dextran (LRD) in a 4:1 ratio.
Amino acid sequence of XSek-1. (A) The deduced amino acid sequence of XSek-1 is shown aligned with mouse Sek-1. The signal peptide and transmembrane domain are indicated by double underlining, and the kinase domain by single underlining. The C termini of the truncated proteins encoded by the constructs used in this study are indicated with an arrow. (B) The percentage amino acid sequence identity between Eph-related RTKs most closely related to XSek-1 is tabulated. These genes are more closely related to each other than they are to any other published Eph-related RTKs.
Amino acid sequence of XSek-1. (A) The deduced amino acid sequence of XSek-1 is shown aligned with mouse Sek-1. The signal peptide and transmembrane domain are indicated by double underlining, and the kinase domain by single underlining. The C termini of the truncated proteins encoded by the constructs used in this study are indicated with an arrow. (B) The percentage amino acid sequence identity between Eph-related RTKs most closely related to XSek-1 is tabulated. These genes are more closely related to each other than they are to any other published Eph-related RTKs.
Microinjection of Xenopus and zebrafish embryos
Fertilised and dejellied (Smith and Slack, 1983) Xenopus eggs were equilibrated with 4% Ficoll in 3/4 NAM at least 5 minutes before injection. 5-10 nl of RNA/LRD mixture was microinjected into fertilised Xenopus embryos (Moon and Christian, 1989), either at the one cell stage, or into one cell at the two cell stage. In pilot experiments we found it necessary to inject 5-10 ng RNA in order to obtain phenotypes, and that at mid-neurula stages the concentration of residual injected RNA was similar to that of endogenous XSek-1 transcripts. Embryos were gradually equilibrated with 1/10 NAM 6 hours after injection (prior to gastrulation), and harvested at various neurula stages by manually removing the vitelline membrane and fixing overnight in MEMFA.
Zebrafish embryos were produced by natural spawning and microinjected at the two cell stage using a glass capillary needle attached to a Picospritzer. Approximately 0.3-1 ng of RNA was injected and embryos were allowed to develop to various neurula stages before fixation in 4% paraformaldehyde in PBS. Manual dechorionation was carried out before in situ hybridisation or immunocytochemical analysis.
Whole-mount in situ hybridisation and immunocytochemistry
The analysis of Xenopus embryos by whole-mount in situ hybridisation was carried out essentially as described by Harland (1991), except that the post-hybridisation RNAse treatment was omitted. The XKrox-20 probe has been described previously (Bradley et al., 1992). The XSek-1 probe is a 2.6 kb fragment including extracellular domain and kinase domain sequences. An identical pattern is observed using a probe against 3′ untranslated sequences, but this short probe gave weaker signals. Whole mount in situ hybridisation and immunocyto-chemistry of zebrafish embryos was performed as described (Xu et al., 1994), using krx20 (Oxtoby and Jowett, 1993) or rtk1 (Xu et al., 1994) probes, or anti-Pax6 (Macdonald et al., 1994) antibody. In order to examine segmental expression patterns in detail, Xenopus or zebrafish embryos were partially dissected such that the hindbrain could be mounted under a coverslip, either in a dorsal or lateral view.
Retrograde tracing of reticulospinal neurons
Retrograde tracing with LRD was carried out as described by Hill et al. (1995) except that low melting temperature agarose in 10% Hank’s saline was used to immobilise the embryos.
RESULTS
Cloning of Xenopus and zebrafish orthologues of Sek-1
We cloned the Xenopus orthologue of Sek-1 by screening a neurula-stage cDNA library with a probe corresponding to the tyrosine kinase domain of mouse Sek-1. Sequence analysis revealed an open reading frame with high sequence identity to mouse Sek-1 (Fig. 1A). Sequence comparisons also showed a strong identity with Cek8, the chick orthologue of Sek-1, but less similarity with Ehk-1, Ehk-2 and Hek, the Eph-related genes most closely related to Sek-1 (Fig. 1B). Taken together with the similarity between the expression pattern of this gene in Xenopus and Sek-1 in mouse (see below), these data indicate that we have cloned the Xenopus orthologue of Sek-1, which we designate XSek-1. While our functional studies were in progress, a very similar cDNA, named Pag, was reported (Winning and Sargent, 1994); Pag and XSek-1 are probably polymorphic alleles of the same gene, perhaps reflecting the tetraploidy of the Xenopus laevis genome.
In earlier studies, zebrafish cDNA clones related to Sek-1 were isolated by screening of a neurula-stage library with sequences from the tyrosine kinase domain of XSek-1 (Xu et al., 1994). As previously described, partial sequencing and the expression pattern indicates that one of these, named rtk1, corresponds to the zebrafish orthologue of Sek-1.
Developmental regulation of Sek-1/rtk1 in the Xenopus and zebrafish hindbrain
To characterise the expression patterns of Sek-1 orthologues in Xenopus and zebrafish embryos we carried out whole mount in situ hybridisation analysis. Many, but not all, aspects of Sek-
1 expression found in the mouse (Nieto et al., 1992) are conserved in Xenopus, including expression in early mesoderm and the notochord (not shown), r3, r5, and neural crest adjacent to r5 (Fig. 2A-F), the otic placode (Fig. 2D), the forebrain and the olfactory placode (Fig. 2A,B). In the hindbrain, XSek-1 transcripts are first detected at stage 14 in presumptive r3, and by stage 14.5 have been up-regulated in presumptive r5 (Fig. 2A). Expression is detected in neural crest adjacent to pre-r5 just prior to expression in pre-r5 itself, as seen in embryos with a slight asynchrony in development between left and right (Fig. 2A). At stages 15-16, XSek-1 expression is not sharply restricted and low levels of transcripts are detected in presumptive r4 (Fig. 2B,C). By stage 20 the expression domains have become precisely restricted to r3 and r5, with a lower-level domain in r2 (Fig. 2D,E). Neural crest cells expressing XSek-1 migrate into the third branchial arch (Fig. 2D). Whereas XSek-1 transcripts persist in r3 and r5, expression in r2 is transient, and a third, more rostral stripe of Sek-1 expression appears in the dorsal part of r1 between stages 31-33 (Fig. 2F).
Expression patterns of XSek-1 and rtk1 in the developing hindbrain. (A-F) The expression pattern of XSek-1 during Xenopus development was analysed by whole mount in situ hybridisation. Photographs were taken of either cleared whole embryos (A,D,F), or of the neural epithelium after mounting under a coverslip (B,C,E). (A) Stage 14.5. (B) Rostral neural epithelium at stage 15. (C) Higher magnification view of hindbrain at stage 15. (D) Stage 20. (E) Hindbrain at stage 20. (F) Stage 33. (G-I) The expression pattern of rtk1 during zebrafish development was analysed by whole mount in situ hybridisation. Photographs from a dorsal view were taken after mounting of the hindbrain under a coverslip. (G) 11.5 h. (H) 17 h. (I) 24 h. r, rhombomere; fb, forebrain; nc, neural crest; ol, olfactory placode; ot, otic placode; p, pronephros. Scale bars, 50 μm.
Expression patterns of XSek-1 and rtk1 in the developing hindbrain. (A-F) The expression pattern of XSek-1 during Xenopus development was analysed by whole mount in situ hybridisation. Photographs were taken of either cleared whole embryos (A,D,F), or of the neural epithelium after mounting under a coverslip (B,C,E). (A) Stage 14.5. (B) Rostral neural epithelium at stage 15. (C) Higher magnification view of hindbrain at stage 15. (D) Stage 20. (E) Hindbrain at stage 20. (F) Stage 33. (G-I) The expression pattern of rtk1 during zebrafish development was analysed by whole mount in situ hybridisation. Photographs from a dorsal view were taken after mounting of the hindbrain under a coverslip. (G) 11.5 h. (H) 17 h. (I) 24 h. r, rhombomere; fb, forebrain; nc, neural crest; ol, olfactory placode; ot, otic placode; p, pronephros. Scale bars, 50 μm.
In zebrafish, rtk1 is expressed in a similar pattern in mesoderm, the hindbrain, otic placode and forebrain as Sek-1 in higher vertebrates (Fig. 2G-I; Macdonald et al., 1994; Xu et al., 1994). In the hindbrain, rtk1 expression is up-regulated in pre-r3 at 10h (10 hours of development; not shown), and then in pre-r5 at 11.5h (Fig. 2G). These expression domains become sharply restricted to r3 and r5 at 14h, and at 17h low levels of transcripts are also detected in r2 (Fig. 2H). Expression in r2 is transient, whereas transcripts continue to be expressed in r3 and r5. A third, narrow stripe of rtk1 expression appears in r1 at 24h (Fig. 2I).
These studies indicate that in Xenopus and zebrafish, as in higher vertebrates, expression of Sek-1/rtk1 occurs in r3, r5 and transiently at lower levels in r2. However, the later expression in r1 in zebrafish and Xenopus does not occur in the mouse.
Strategy to interfere with Sek-1 function
We investigated the function of Sek-1 by taking a dominant negative approach analogous to that used to study the role of signalling through the FGF, activin and BMP4 receptors in the Xenopus embryo (Amaya et al., 1991; Hemmati-Brivanlou and Melton, 1992; Graff et al., 1994). Receptor kinases are activated by a ligand-induced dimerisation of receptor leading to trans-phosphorylation and activation of the intracellular catalytic domain (Ullrich and Schlessinger, 1990). Activation can therefore be disrupted by the over-expression of truncated receptor, comprising the extracellular and transmembrane domains, but lacking kinase function. Upon binding of the ligand this truncated receptor is capable of dimerising with the endogenous receptor to form a complex that, because of the absence of kinase function, cannot activate the catalytic domain of the endogenous protein (Amaya et al., 1991; Ueno et al., 1991, 1992).
To interfere with Sek-1 function, we cloned sequences encoding truncated mouse and Xenopus Sek-1 (see Fig. 1) into a modified version of the pSP64T vector (Krieg and Melton, 1984), from which capped RNA can be transcribed. A construct to provide a negative control for any non-specific effects of RNA injection was generated by introducing a frame shift mutation 3′ to the signal peptide sequence of truncated Sek-1. In vitro transcription and translation confirmed that polypeptides of the predicted length are encoded by these sequences (data not shown). These reagents allowed us to ascertain whether there is any difference in the effects of expressing, in Xenopus embryos, truncated mouse Sek-1, compared with the homologous truncated XSek-1. Full-length rtk1 sequences have not yet been obtained, so expression of homologous truncated receptor has not been carried out in the zebrafish embryo.
Truncated Sek-1 disrupts segmental gene expression in the Xenopus hindbrain
RNA encoding truncated mouse or Xenopus Sek-1 was microinjected together with lineage tracer into Xenopus embryos, either at the 1 cell stage, or into one cell at the 2 cell stage; the latter injections provide an internal comparison of any phenotype generated because the injected RNA is present only in the left or right half of the embryo. Detection of this RNA revealed a uniform distribution at higher levels than endogenous Sek-1 RNA within the injected half at early-mid neurula stages (not shown). Embryos were fixed at neurula stages and analysed by in situ hybridisation with XKrox-20, a molecular marker of r3 and r5 identity.
During normal development, XKrox-20 expression is up-regulated first in pre-r3 at stage 14 and then in pre-r5 at stage 14.5, and by stage 17 has become sharply restricted to r3 and r5 (Bradley et al., 1992). Following the injection of RNA encoding truncated Sek-1 receptor into Xenopus embryos, an altered pattern of XKrox-20 gene expression was observed in 12% (15/124) of embryos analysed at neurula stages. We observed similar phenotypes when RNA encoding truncated Xenopus (Fig. 3B,F) or mouse (Fig. 3C-E) Sek-1 was injected, and in embryos injected at the 2 cell stage, the affected side always correlated with the presence of co-injected lineage tracer. At stage 14, the pre-r3 expression domain of XKrox-20 forms a narrow band (Fig. 3A), whereas after injection of RNA encoding truncated receptor some expressing cells encroach into the adjacent non-expressing territory (Fig. 3B). At stage 15, after the up-regulation of pre-r5 expression, a similar encroachment of cells expressing XKrox-20 into presumptive even-numbered rhombomeres is observed in the injected, but not the control half embryo (Fig. 3C). At later stages, coherent groups of cells expressing XKrox-20 are observed extending from r3/r5 into adjacent even-numbered territory (Fig. 3D-F).
Effect of truncated Sek-1 on XKrox-20 gene expression in the Xenopus hindbrain. RNA encoding truncated Sek-1 was microinjected into 1 cell of 2 cell Xenopus embryos, so that one half of the embryo received injected RNA. The embryos were allowed to develop to various neurula stages and fixed. In situ hybridisation was then carried out to analyse the expression pattern of XKrox-20, a molecular marker of r3 and r5. Photographs were taken of the mounted hindbrain. (A) Uninjected half of stage 14 embryo. (B) Injected half of stage 14 embryo. (C) Stage 15 embryo; the injected half is on the left. (D-F) Stage 16 embryos, with injected RNA present in the left (D) or right (E,F) half. r, rhombomere. The arrowheads indicate XKrox-20-expressing cells in even-numbered rhombomeres. Scale bars, 50 μm.
Effect of truncated Sek-1 on XKrox-20 gene expression in the Xenopus hindbrain. RNA encoding truncated Sek-1 was microinjected into 1 cell of 2 cell Xenopus embryos, so that one half of the embryo received injected RNA. The embryos were allowed to develop to various neurula stages and fixed. In situ hybridisation was then carried out to analyse the expression pattern of XKrox-20, a molecular marker of r3 and r5. Photographs were taken of the mounted hindbrain. (A) Uninjected half of stage 14 embryo. (B) Injected half of stage 14 embryo. (C) Stage 15 embryo; the injected half is on the left. (D-F) Stage 16 embryos, with injected RNA present in the left (D) or right (E,F) half. r, rhombomere. The arrowheads indicate XKrox-20-expressing cells in even-numbered rhombomeres. Scale bars, 50 μm.
Truncated Sek-1 disrupts segmental gene expression in the zebrafish hindbrain
Although it was unclear why, despite the uniform overexpression of truncated Sek-1, disruption occurred only in a low proportion of embryos, these results suggest that Sek-1 function is required for the segmental restriction of gene expression to r3/r5 in Xenopus. To examine whether this function is conserved in a different vertebrate class, we carried out analogous experiments in the zebrafish. RNA encoding truncated Xenopus or mouse Sek-1 was microinjected into 1 cell of the 2 cell zebrafish embryo. At early blastula stages, cleavage is initially partial so RNAs could diffuse between cells, but in pilot experiments we nevertheless observed a mosaic inheritance of the injected RNA (data not shown). Injected embryos were allowed to develop to various neurula stages, fixed and then analysed by in situ hybridisation with rtk1 or krx20 probes, as markers of r3/r5. The rtk1 probe corresponds to 3′ untranslated sequences not present in the injected Sek-1 RNA.
At early neurula stages (14h), shortly after the onset of rtk1 and krx20 expression, >50% of embryos expressing truncated Sek-1 or XSek-1 receptor had major disruptions to the spatial restriction of krx20 expression. In contrast to normal embryos, in which krx20 expression occurs in parallel, narrow bands (Fig. 4A; Oxtoby and Jowett, 1993), in embryos expressing truncated receptor these domains are irregular in shape, and often overlap (Fig. 4B,C). Similar severe disruptions are also observed when rtk1 expression is used as a marker (not shown). We observed no difference between truncated mouse or Xenopus Sek-1 in their effects at this or later stages. Analysis of 18-24h embryos, when in normal development r3/r5 gene expression has become sharply restricted (Fig. 4D), revealed disruptions to rtk1 or krx20 expression in 55% (41/74) of embryos, and these fell into two classes of phenotypes (Fig. 4E-L). In 65% (27/41) of the affected embryos, cells were observed extending from an apparently normal r3/r5 expression domain into the adjacent r2 or r6 territory (Fig. 4E,F). 35% (14/41) of the affected embryos showed a more drastic phenotype, with an apparent fusion between r3 and r5, as seen in dorsal views (Fig. 4G,H,J) and sections (Fig. 4I). In many of these embryos the shape of r3 or r5 is altered, for example with one side narrower than normal (Fig. 4H,I). In all cases, cells ectopically expressing r3/r5 markers were not isolated, but formed coherent and sharply restricted populations intruding into even-numbered territory. Lateral views revealed that in both classes of embryos, the ectopic gene expression and fusions are dorsally located, and non-expressing cells are present ventrally in presumptive r4 (Fig. 4K,L). We therefore infer that r4 is still present, but no molecular markers are currently available to assess this directly. In summary, expression of truncated Sek-1 in zebrafish leads to a similar disruption of the spatial restriction of r3/r5 markers as in Xenopus, but with more severe phenotypes and in a higher proportion of embryos.
Effect of truncated Sek-1 on krx20 and rtk1 expression in the zebrafish hindbrain. RNA encoding truncated Sek-1 was microinjected into 1 cell of zebrafish embryos at the 2 cell stage. Embryos were allowed to develop to neurula stages, fixed and either krx20 or rtk1 expression analysed by in situ hybridisation. Photographs were taken either of the hindbrain from a dorsal (A-H,J) or lateral view (K,L), or after sectioning in the coronal plane (I). krx20 expression was analysed in (A) uninjected or (B,C) injected embryos fixed at 14 hours of development. rtk1 expression was analysed in uninjected (D) or injected (E-K) embryos analysed at 24 hours of development. (L) krx20 expression in injected 18h embryo. The arrowheads indicate cells expressing krx20 or rtk1 in even-numbered rhombomeres. Scale bars 50 μm.
Effect of truncated Sek-1 on krx20 and rtk1 expression in the zebrafish hindbrain. RNA encoding truncated Sek-1 was microinjected into 1 cell of zebrafish embryos at the 2 cell stage. Embryos were allowed to develop to neurula stages, fixed and either krx20 or rtk1 expression analysed by in situ hybridisation. Photographs were taken either of the hindbrain from a dorsal (A-H,J) or lateral view (K,L), or after sectioning in the coronal plane (I). krx20 expression was analysed in (A) uninjected or (B,C) injected embryos fixed at 14 hours of development. rtk1 expression was analysed in uninjected (D) or injected (E-K) embryos analysed at 24 hours of development. (L) krx20 expression in injected 18h embryo. The arrowheads indicate cells expressing krx20 or rtk1 in even-numbered rhombomeres. Scale bars 50 μm.
If Sek-1/rtk1 has any adhesive role, it is possible that the disruption was due to the ectopic expression of extracellular domain rather than inhibition of signal transduction. However, no effects on segmental gene expression occurred in >300 zebrafish embryos injected with RNA encoding full-length Sek-1. This enabled us to test whether expression of full-length Sek-1 could rescue the effects of truncated Sek-1. We found that co-injection of increasing amounts of RNA encoding fulllength Sek-1 lead to a progressive suppression of the disruption to the segmental restriction of gene expression (Table 1). These data indicate that Sek-1 is capable of homodimerisation and confirm that the disruption caused by truncated Sek-1 is due to a requirement for the cytoplasmic domain.
Disruption of boundary formation and neuronal organisation by truncated Sek-1
In the chick, morphological boundaries at the interface of odd and even numbered rhombomeres exhibit several distinctive cellular and antigenic properties (Lumsden and Keynes, 1989; Heyman et al., 1993). We found that, in addition to uniform expression up to the r1/r2 boundary, zebrafish Pax6 (PaxA; Krauss et al., 1991; Puschel et al., 1992) protein is expressed at higher levels at rhombomere boundaries in 24h embryos (Fig. 5A). We analysed whether expression of truncated Sek-1 receptor affected the expression pattern of this boundary marker. In 30% of embryos, lower levels of Pax6 expression occurred in certain rhombomere boundaries, for example at the r2/r3 (Fig. 5B) and r5/r6 (Fig. 5C) boundary. In some of these embryos, for example in rostral part of hindbrain shown in Fig. 5C, certain boundaries are distorted, perhaps reflecting the altered shape of rhombomeres detected by rtk1 and krx20 expression (Fig. 4E-L).
Effect of truncated Sek-1 on Pax6 expression at rhombomere boundaries in the zebrafish. Immunocytochemistry was carried out to examine Pax6 expression in (A) uninjected 24h embryos or (B,C) 24h embryos injected with RNA encoding truncated Sek-1. The black arrowheads indicate rhombomere boundaries, all of which, in control embryos, express Pax6 at higher levels than within rhombomeres. The white arrowheads indicate rhombomere boundaries in injected embryos in which Pax6 expression appears deficient. Scale bars, 50 μm.
Effect of truncated Sek-1 on Pax6 expression at rhombomere boundaries in the zebrafish. Immunocytochemistry was carried out to examine Pax6 expression in (A) uninjected 24h embryos or (B,C) 24h embryos injected with RNA encoding truncated Sek-1. The black arrowheads indicate rhombomere boundaries, all of which, in control embryos, express Pax6 at higher levels than within rhombomeres. The white arrowheads indicate rhombomere boundaries in injected embryos in which Pax6 expression appears deficient. Scale bars, 50 μm.
Previous studies have shown a segmental organisation of identified reticulospinal neurons in the zebrafish hindbrain that can be revealed by retrograde labelling from the spinal cord (Metcalfe et al., 1986; Fig. 6A). These neurons are therefore amenable markers of regional specification in the hindbrain (Hill et al., 1995). We examined whether expression of truncated Sek-1 had effects on the organisation of reticulospinal neurons by using lysinated rhodamine dextran for retrograde labelling. Many embryos expressing truncated Sek-1 had the normal number and organisation of reticulospinal neurons indicating that, as suggested by the gene expression data, rhombomere specification has not been altered. However, in some embryos, the spacing of these neurons within rhombomeres is abnormal; for example, r5 neurons (MiD2cm and MiD2cl; Metcalfe et al., 1986) which are close together in the normal hindbrain (Fig. 6A) are further apart in some embryos expressing truncated Sek-1 (Fig. 6B). 10% of injected embryos had a duplication of the Mauthner neuron in r4, as identified by its location and contralateral projection (Fig. 6C).
Effect of truncated Sek-1 on reticulospinal neurons in the zebrafish embryo. Reticulospinal neurons were revealed in 3 day zebrafish embryos by retrograde labelling from the spinal cord using LRD (A) Uninjected embryo. (B,C) Embryos injected with RNA encoding truncated Sek-1. The small arrowheads in A and B indicate the locations of corresponding pairs of reticulospinal neurons in r5 and r6. The spacing of these neurons is altered in the injected embryo shown in B. The small arrowheads in C indicate the axons of the duplicated Mauthner neurons in r4. The large arrows labelled ‘m’ indicate the Mauthner neuron. Scale bar, 50 μm.
Effect of truncated Sek-1 on reticulospinal neurons in the zebrafish embryo. Reticulospinal neurons were revealed in 3 day zebrafish embryos by retrograde labelling from the spinal cord using LRD (A) Uninjected embryo. (B,C) Embryos injected with RNA encoding truncated Sek-1. The small arrowheads in A and B indicate the locations of corresponding pairs of reticulospinal neurons in r5 and r6. The spacing of these neurons is altered in the injected embryo shown in B. The small arrowheads in C indicate the axons of the duplicated Mauthner neurons in r4. The large arrows labelled ‘m’ indicate the Mauthner neuron. Scale bar, 50 μm.
DISCUSSION
It was possible that the expression of Sek-1 in r3 and r5 correlated with a role in establishing or maintaining the segmental identity of cells within these rhombomeres. However, our data argue against such a role, since we do not observe a deficiency in r3/r5 after injection of RNA encoding truncated Sek-1. Rather, we find an ectopic expression of markers of r3/r5 in the Xenopus and zebrafish hindbrain. Unlike the situation in normal development, cells expressing Krox-20 or Sek-1/rtk1 are found in r2, r4 or r6, and in a high proportion of injected zebrafish embryos an apparent fusion occurs between r3 and r5. These ectopic cells occur in a variable pattern in a coherent group co-extensive with r3/r5, and might underlie the distortions of rhombomere boundaries and altered spacing of reticulospinal neurons observed at later stages. To interpret these results, we will first discuss the potential effects of truncated Sek-1, and then consider what mechanisms could underlie the restriction of segmental gene expression.
Specificity of truncated Sek-1
Our strategy is based upon the assumption that, as for other RTKs, activation of Sek-1 occurs through a ligand-induced dimerisation. Although this has not been directly shown for Eph-related RTKs, in cells expressing a fusion protein of EGF receptor ligand binding domain and Elk receptor kinase domain, the latter is phosphorylated upon EGF treatment (Lhotak and Pawson, 1993). Therefore the kinase domain of this Eph-related RTK can be activated by dimerisation. Furthermore, whereas soluble Elk receptor ligand is inactive, a dimerised form of this activates Elk, suggesting that activation involves receptor homodimerisation (Davis et al., 1994). Our finding that co-injection of RNA encoding full-length Sek-1 rescues the phenotype caused by truncated Sek-1 indicates that these can bind to each other and thus provides further evidence for homodimerisation.
The use of truncated receptors has provided important insights into the roles of receptor kinases in Xenopus mesoderm patterning (Amaya et al., 1991; Hemmati-Brivanlou and Melton, 1992; Graff et al., 1994), but have revealed some potential limitations to this approach. In particular, the truncated receptor may interfere with the function of several endogenous receptors by heterodimerisation (Ueno et al., 1992). It is not known whether Eph-related RTKs can heterodimerise, such that truncated Sek-1 could interfere with the function of other family members; the rescuing effects of fulllength Sek-1 do not address this since it could compete for binding of truncated Sek-1 to a heterologous partner. If truncated Sek-1 does inhibit the function of several RTKs the effects we observe could involve related genes co-expressed with Sek-1, such as Sek-3/nuk, Sek-4, and zebrafish rtk3 (Becker et al., 1994; Henkemeyer et al., 1994; Q. X. and N. H., unpublished observations). However, these genes are expressed after segmentation, and no Eph-related RTKs have been found that are co-expressed with Sek-1 in pre-r3/r5, when we first observe disrupted gene expression. If receptor heterodimerisation occurs, less severe effects may result from disruption of the Sek-1 gene compared with the use of a dominant negative approach. Since the phenotype correlates with Sek-1 expression, for simplicity we will not refer below to the possibility of effects on multiple receptors.
Potential mechanisms restricting segmental gene expression
Whereas regional specification occurs prior to segmentation in the chick hindbrain (Guthrie et al., 1992), the progeny of an individual cell marked at this stage disperse considerably, and often contribute to two adjacent rhombomeres (Fraser et al., 1990). Since Krox-20 transcripts are up-regulated in pre-r3 and -r5 at this stage (Nieto et al., 1991), it seems that expression does not correlate with an irrevocable commitment to an r3/r5 identity. Although after segmentation, most clonal progeny are restricted to a single rhombomere (Fraser et al., 1990), in 10-20% of clones some cells cross rhombomere boundaries (Birgbauer and Fraser, 1994). Thus the observation of sharp r3/r5 domains of Krox-20 expression implies that cells are not committed even after segmentation. It therefore seems that there is a dynamic regulation of cell identity. We suggest that restricted domains of Krox-20 gene expression might be maintained by local interactions acting in a community effect (Gurdon, 1988; Gurdon et al., 1993), in which as intermingling occurs between presumptive or definitive rhombomeres, cells switch their segmental identity to that of their neighbours. It is also possible that rather than switching identity, cell death occurs. However, cell death is unlikely to account for the restriction of gene expression prior to segmentation, when individual cells can contribute many progeny to each of two rhombomeres.
The partial restriction of cell movement across rhombomere boundaries (Fraser et al., 1990; Birgbauer and Fraser, 1994) is also likely to contribute to the formation of sharp segmental domains of gene expression. Transplantation experiments in the chick suggest that this restriction involves cellular properties, perhaps adhesion, that to a first approximation alternate between rhombomeres (Guthrie et al., 1993) and might lead to a sorting of cells with odd and even segmental identity. Taken together, these data suggest roles of both identity switching and constraints on cell mixing, but the relative contribution of these mechanisms is not known and may vary during the establishment of segments. Current data are consistent with a critical role of cell identity switching in the maintenance of gene expression domains, with cell adhesion serving to sharpen and stabilise the pattern upon segmentation. Indirect support for this is provided by the phenotype of mice in which r3 and r5 are missing due to inactivation of the Krox-20 gene (Schneider-Maunoury et al., 1993). The juxtaposed r2/r4/r6 in these mutants are predicted to have greater intermingling than odd and even rhombomeres (Guthrie et al., 1993), yet r4-specific gene expression is as sharp as in normal embryos. However, these studies do not rule out the possibility that there is a partial restriction of cell movement at early stages, which has a critical role.
Disruption of segmentally restricted gene expression by truncated Sek-1
We suggest three possible mechanisms by which expression of truncated Sek-1 could lead to cells expressing Krox-20/Sek-1 in even-numbered rhombomeres: a switch of r2/r4/r6 cells to an r3/r5 identity; an increased mixing of cells between odd and even rhombomeres; or a block in the switching of r3/r5 cells that have intermingled into r2/r4/r6. These are discussed below.
According to the first hypothesis, expression of truncated Sek-1 has lead to a de novo up-regulation of Krox-20/Sek-1 expression in even-numbered rhombomeres. Such a mechanism requires that truncated Sek-1 has not disrupted Sek-1 function, but rather an RTK expressed in r2/r4/r6. However, this explanation seems very unlikely since the cells ectopically expressing r3/r5 markers are always found contiguous with r3 or r5. Although ectopic cells with r3/r5 identity might preferentially adhere to r3/r5 during cell mixing, this intermingling would have to be very extensive and rapid to account for the complete absence of isolated ectopic cells in even-numbered rhombomeres; this is especially unlikely in Xenopus, in which little intermingling occurs (Wetts and Fraser, 1989). Furthermore, this hypothesis does not account for the variability in the disrupted pattern of gene expression or the difference in the proportion of affected embryos in zebrafish and Xenopus.
The second hypothesis proposes that presumptive r3/r5 cells are present in r2/r4/r6 due to increased mixing between odd and even rhombomeres. Indeed, since ligands for Eph-related receptors are membrane bound (Bartley et al., 1994; Beckmann et al., 1994; Cheng and Flanagan, 1994; Davis et al., 1994), they could mediate adhesive interactions, and the r3/r5 expression of Sek-1 thus contributes to the adhesive differences between odd-and even-numbered segments. Widespread expression of the extracellular domain of Sek-1 could therefore lessen the adhesive difference between r3/r5 and adjacent rhombomeres, leading to increased mixing of these cell populations. However, we find injection of RNA encoding fulllength Sek-1 has no effect on the segmental restriction of gene expression, suggesting that the effect of truncated receptor is not due to ectopic expression of the extracellular domain. Indeed, full-length Sek-1 rescues the effect of truncated Sek-1, indicating that if this RTK does have a role in restricting cell movement from odd to even rhombomeres, it requires the kinase domain. Thus, rather than mediating a passive adhesion, any role of Sek-1 in lineage restriction might involve an active process, such as a contact dependent repulsion of r3/r5 cells by r2/r4/r6 cells.
According to the third hypothesis, truncated Sek-1 protein interferes with signal transduction and disrupts the ability of presumptive r3/r5 cells to switch segmental identity if, during intermingling, they cross into presumptive r2/r4/r6. The fusion of r3 and r5 observed in many zebrafish embryos may be a consequence of pre-r3/r5 cells encroaching sufficiently into pre-r4 to bridge the normal 3-4 cell wide gap between these oddnumbered rhombomeres in the early neural plate. Since cells from r3 and r5 will mix with each other relatively freely (Guthrie et al., 1993), touching of these rhombomeres could lead to an irreversible fusion. This model suggests that during normal development, activation of Sek-1 receptor occurs in any cells that move from odd-to even-numbered territory and is required for a switch in phenotype to that of the local community. The switch in phenotype is accompanied by the down-regulation of Sek-1/rtk1 and Krox-20 transcripts, which must occur rapidly to account for the sharp expression domains. This model predicts that the ligand for Sek-1 is expressed in even-numbered rhombomeres and that the signalling interactions are short-range. Indeed, ligands for Ephrelated RTKs are active only when membrane-bound and not in soluble form (Davis et al., 1994). Recently, a ligand, Elf-1, has been identified that binds Sek-1 and another Eph-related RTK, Mek-4 (Cheng and Flanagan, 1994). However, expression of Elf-1 does not occur within, or adjacent to r3/r5, and it is therefore unlikely that Elf-1 activates Sek-1 in the hindbrain.
An important question arising from this model is how boundaries are stabilised, since at this interface pre-r3/r5 cells are in contact with pre-r2/r4/r6. One possibility is that a switch from odd to even identity requires contact with more than a critical ratio of pre-r2/r4/r6 cells relative to r3/r5 cells. It may be significant that after segmentation in the chick hindbrain a distinct population of cells is formed at boundaries, across which there is less cell contact compared to within rhombomeres (Martinez et al., 1992), perhaps due to larger intercellular spaces (Lumsden and Keynes, 1989; Heyman et al., 1993). It is possible that this relative lack of cell contact might stabilise domains of gene expression after segmentation, and thus it is intriguing that we find truncated Sek-1 leads to a deficiency in the higher-level expression of Pax6 in rhombomere boundary cells in the zebrafish. Studies in the chick indicate that boundaries form at the interface between any combination of odd-and even-numbered rhombomeres (Guthrie and Lumsden, 1991), so boundary cells may be induced by local interactions between these populations. We speculate that the decrease in Pax6 expression in some boundary cells may reflect a role of Sek-1 in such interactions between adjacent rhombomeres. If boundary cells stabilise gene expression domains, might disruption of their formation underlie the ectopic expression of r3/r5 markers in the presence of truncated Sek-1? Although this cannot be ruled out, current data do not support this possibility since we observe disruptions at early stages whereas boundary cells have only been detected after segmentation. The effects of truncated Sek-1 on boundary cells may therefore reflect a distinct, later role.
Variations in the extent and pattern of disruption
Our data raise the questions as to why the extent of disruption and proportion of affected embryos is low in Xenopus despite the uniform distribution of injected RNA, why it is lower than in zebrafish, and why in both species the patterns of disruption are variable. These may relate to the extent and variability of cell mixing normally occurring in the hindbrain, and are consistent with models in which ectopic gene expression derives from r3/r5. Clonal analyses in the zebrafish have shown that during convergent extension there is considerable intermingling of cells along the A-P axis (Kimmel et al., 1994). This occurs because the progeny of a cell division often lie along the A-P axis, and are subsequently separated by the intercalation of other cells. Although the dispersal of clones has not been followed after the 16th cell division (8-9 h), convergent extension is still occurring after the onset of krx20 expression (Oxtoby and Jowett, 1993), so substantial intermingling along the A-P axis seems likely to occur. As a consequence, blocking of the restriction of cell movement between presumptive rhombomeres or of cell identity switching will frequently lead to severe disruption of spatially restricted gene expression, and since the patterns of cell division and intercalation are stochastic (Kimmel et al., 1994) the extent of this disruption is variable. In contrast, little cell intermingling occurs in Xenopus, even during the major morphogenetic movements between blastula and mid-neurula stages (Wetts and Fraser, 1989). Moreover, only one cell division occurs in the neural epithelium between stages 13 and 16 (Hartenstein, 1989), the period when Krox-20/Sek-1 expression is established and sharpened. There is therefore little opportunity for the dispersal of clonally-related cells along the A-P axis by division and intercalation. As a consequence, movement of cells between presumptive rhombomeres and a switching of segmental identity is likely to occur infrequently in Xenopus.
If substantial cell mixing is occurring in the zebrafish hindbrain, why do we find that cells expressing r3/r5 markers in an ectopic location always form a coherent group, rather than being interspersed with non-expressing cells? One possibility is that due to turnover of the injected RNA encoding truncated Sek-1 there is only a transient inhibition of lineage restriction or of cell identity switching which causes the severe scrambling seen in the 12h embryo. From this the coherent arrangement of r3/r5 cells seen at 18-24h could then emerge by a sorting of odd-and even-numbered cells and/or because ectopic communities, but not isolated cells, can maintain r3/r5 phenotype. According to this, the variable and abnormal shapes of r3/r5 seen at later stages may reflect the stochastic nature of the initial scrambling of cells with presumptive odd-and even-numbered identity, due to variability in planes of cell division (Kimmel et al., 1994).
Concluding remarks
Our results indicate that Sek-1 function is required for the restriction of Krox-20 and Sek-1 gene expression to r3/r5, and suggest a role either in a dynamic regulation of cell identity in a segmental community effect or in a restriction of cell movement from odd to even presumptive segments. The relative contribution of these mechanisms to the establishment of segmental gene expression domains is unknown, and further understanding of the role of Sek-1 will therefore require cell lineage analysis and transplantation experiments.
REFERENCES
Note added in proof
The nucleotide sequence of XSek-1 will appear in the EMBL, GenBank and DDBJ Nucleotide Sequence Databases under the accession number X91191.