ABSTRACT
In zebrafish, individual primary motoneurons can be uniquely identified by their characteristic cell body positions and axonal projection patterns. The fate of individual primary motoneurons remains plastic until just prior to axogenesis when they become committed to particular identities. We find that distinct primary motoneurons express particular combinations of LIM homeobox genes. Expression precedes axogenesis as well as commitment, suggesting that LIM homeobox genes may contribute to the specification of motoneuronal fates. By transplanting them to new spinal cord positions, we demonstrate that primary motoneurons can initiate a new program of LIM homeobox gene expression, as well as the morphological features appropriate for the new position. We conclude that the patterned distribution of different primary motoneuronal types within the zebrafish spinal cord follows the patterned expression of LIM homeobox genes, and that this reflects a highly resolved system of positional information controlling gene transcription.
INTRODUCTION
The development of a functional vertebrate nervous system requires elaboration of a large number of diverse cell types in highly specific locations. We have been studying the development of a small set of identified neurons, primary motoneurons, in zebrafish to understand how neuronal cell fates become specified. Zebrafish primary motoneurons can be identified by their cell body positions and axonal trajectories (Eisen et al., 1986, 1990). They are serially distributed in bilateral clusters that correspond to the adjacent, segmentally arranged somites. Each cluster is composed of three distinct primary motoneurons, RoP, MiP and CaP with a fourth primary motoneuron, VaP, variably present. Every cell body assumes a stereotyped position within the ventral spinal cord and the axons of a segmental group exit the spinal cord along a common pathway, forming a ventral root. The axons diverge a short distance from the spinal cord to innervate distinct motoneuron-specific regions of the corresponding axial musculature.
Transplantation experiments showed that primary motoneurons become committed to a particular fate after their final division but before axogenesis (Eisen, 1991). Primary motoneurons are born beginning at about 10 hours (postfertilization at 28.5°C) (Myers et al., 1986; Kimmel et al., 1994). Axogenesis begins at about 17 hours depending on the axial level, but the cells can be recognized by their cell body positions several hours before axonal outgrowth. When transplanted to heterotopic locations within the ventral spinal cord just before axogenesis, primary motoneurons exhibited axonal trajectories and, thus, fates appropriate for their original positions. If, however, a primary motoneuron was transplanted to a new location at an earlier time relative to axogenesis, it adopted a primary motoneuronal fate appropriate for its new position as indicated by its axonal trajectory (Eisen, 1991). These observations suggest that, after a period of plasticity, primary motoneurons become committed to a particular fate that is determined by the local environment.
What are the molecular mechanisms that underlie specification to particular neuronal cell fates? To address this question we have begun to identify genes expressed in primary motoneurons prior to their morphological differentiation. Among these are several members of the homeobox subclass of the LIM gene family from zebrafish, including homologs of the genes described by Tsuchida et al. (1994), which are expressed in chick motoneurons. Previous work has shown that one of these genes, islet1, is among the earliest known genes expressed in motoneurons of rat and chick and primary motoneurons of zebrafish (Thor et al., 1991; Ericson et al., 1992; Korzh et al., 1993; Inoue et al., 1994). Genetic studies have shown that LIM homeobox genes are important for development, including the specification of neuronal fates. For example, in Caenorhabditis elegans, mec-3 is necessary for differentiation of mechanosensory neurons (Way and Chalfie, 1988), in Drosophila melanogaster, apterous is required for axonal pathway selection by a subset of interneurons (Lundgren et al., 1995) and, in mice, Lim1 is required for head development (Shawlot and Behringer, 1995).
To begin a detailed analysis of the role of the LIM homeobox gene family in primary motoneuron fate specification, we have isolated several corresponding cDNAs from zebrafish. We show that zebrafish LIM homeobox genes are combinatorially expressed in primary motoneurons and that distinct primary motoneurons express different combinations of these genes. We show that expression of these genes in zebrafish primary motoneurons begins well before overt differentiation and that the patterns are highly dynamic. The resolution of the RNA expression patterns precedes the commitment of a cell to a particular motoneuronal fate. Finally, using cell transplantation experiments, we demonstrate that the expression of a LIM homeobox gene is determined by the position of a primary motoneuron within the spinal cord and, furthermore, that commitment of a primary motoneuron to a particular fate is correlated with a distinct pattern of LIM homeobox gene expression.
MATERIALS AND METHODS
Isolation of zebrafish LIM homeobox genes
Approximately 1×106 plaques of a 20-28 hour lambda ZAP-II cDNA library (gift of Robert Riggleman, Kathryn Helde and David Grunwald) were screened at low stringency using a fragment encompassing the rat Islet-1 homeobox (Karlsson et al., 1990) to isolate both zebrafish islet1 and islet2 clones. cDNA inserts were subcloned by helper-phage cotransfection (Stratagene) and mapped with restriction enzymes. Both strands were sequenced by the dideoxy chain termination method using a T7 Sequencing Kit (Pharmacia). Compressions were resolved by the use of Deaza G/A T7 Sequencing Mixes (Pharmacia). DNA sequences and derived amino acid sequences were analysed on a VAX computer using Genius software package (EMBL, Heidelberg).
Zebrafish lim3 genomic sequences, isolated by low-stringency screening of a genomic library (gift of Barbara Jones and Martin Petkovich), were the basis for designing the following primers: L3R2, 5′-TATCGAATTCAAATGTAGTACCTACTC; L1F1, 5′-ATATCC-CGGTATGTGCGG; cR1, 5′-CGAAGAGCTGAAAGCTTTCC; L1F2, 5′-TTCTGGATCGCCACTGGC; cR2, 5′-ATCGCT-GCAGTCTATGG. Primer L3R2 was used to initiate cDNA from total RNA (Westerfield, 1994) from 1-day-old zebrafish embryos. The cDNA was amplified with L1F1 and cR1 primers, and the product reamplified with the nested primers L1F2 and cR2, yielding the predicted lim3 band of 1150 basepairs. This DNA was cloned into the pCR-Script SK (+) vector (Stratagene) to produce cDNA clone pIS8 representing the zebrafish lim3 coding sequence from 127 nucleotides downstream of the predicted AUG start site to 31 nucleotides past the stop codon. This cDNA was used as probe in the experiments reported here. The remaining N-terminal coding sequence was generated by the RACE procedure using the materials and instructions provided by Clonetech.
Embryos
Embryos were obtained from the University of Oregon laboratory colony and raised at 28.5°C in embryo medium (13.7 mM NaCl, 0.54 mM KCl, 1.3 mM CaCl2, 1.0 mM Mg SO4, 0.044 mM KH2PO4, 0.025 mM NaH2PO4, 0.42 mM NaHCO3 [pH 7.2]). Staging was according to somite formation (Kimmel et al., 1995).
Intracellular dye labeling
Individual primary motoneurons were injected with a combination of lysinated rhodamine dextran (3,000 Mr, Molecular Probes) and lysinated FITC dextran (3,000 Mr, Molecular Probes) as described previously (Eisen et al., 1990). Donor embryos for primary motoneuron transplants were labeled by injecting a mixture of lysinated rhodamine dextran and lysinated biotin dextran (10,000 Mr, Molecular Probes) into the yolk of early cleavage stage embryos as described previously (Ho and Kane, 1990). All blastomeres take up dye through cytoplasmic connections with the yolk.
RNA in situ hybridization and photography
RNA probes were prepared by transcribing linearized cDNA clones with T3 or T7 polymerase in the presence of digoxygenin or fluorescein labeling mix (Boehringer Mannheim). In situ hybridization was carried out essentially as described previously (Thisse et al., 1993) except that the probes were not hydrolized. Double RNA in situ hybridization and detection was described previously (Hauptmann and Gerster, 1994). For injection and transplantation experiments, injected labels were histochemically detected, following completion of the hybridization procedure, using anti-fluorescein (Boehringer Mannheim) and anti-biotin (Sigma) antibodies, respectively, conjugated to alkaline phosphatase, which were used with Fast Red substrate (Boehringer Mannheim) to form a red precipitate. Embryos were mounted in 75% glycerol in PBS and photographed with Nomarski optics using a Zeiss Axioplan and a Kodak DCS 400 digital camera. The Fast Red reaction product is fluorescent when illuminated and viewed with a Texas Red filter set, which we used to visualize axon projections and low levels of RNA in double labeling procedures. Digital images were processed on a Power Macintosh using Adobe Photoshop 3.0, assembled with Microsoft Powerpoint and printed with a Tektronix Phaser 440 dye sublimation printer. Image processing consisted of cropping and resizing, contrast enhancement, color correction, and brightness, hue and saturation adjustment. Fluorescence and bright-field images of transplanted cells were recorded prior to fixation with an image intensifier plate connected to a video camera and stored separately on an optical disc using Axovideo software (Myers and Bastiani, 1991; Axon Instuments, Inc.). Processing consisted of addition of images from different focal planes, contrast and image enhancement, pseudocolor addition, and addition of fluorescence and bright-field images.
Primary motoneuron transplants
Transplants were carried out as described previously (Eisen, 1991). Donor and host embryos ranged between the 12.5-somite stage and the 18-somite stage. MiPs were removed from segment levels 6-14 of donor embryos and transplanted to segment levels 5-11 of hosts. Native primary motoneurons were removed from recipient spinal segments prior to the transplantation.
RESULTS
Isolation of LIM homebox genes from zebrafish
We cloned zebrafish LIM homeobox genes by screening libraries with probes generated from rat and Xenopus laevis cDNA clones (Materials and Methods). Isolation of zebrafish islet1 has been reported (Inoue et al., 1994). Relationships to LIM homeobox genes identified in other vertebrates were determined by nucleotide sequence comparison (Dawid et al., 1995). Fig. 1 shows an alignment of the amino acid sequences spanning the LIM domains and homeodomains encoded by the islet1, islet2 and lim3 genes, which we focus on for the present study. The islet1 and islet2 proteins are very similar having a single amino acid difference within the homeodomain, and 94% and 70% amino acid sequence identity in the first and second LIM domains, respectively. The lim3 protein is quite distinct from the islet1 and islet2 proteins, sharing 47% sequence identity within the homeodomain, and 56% and 51% identity within the first and second LIM domains. Zebrafish lim3 is very similar to lim3 of other vertebrates. For example, it has 98% sequence identity in the homeodomain and 95% identity in the LIM domains compared to its X. laevis ortholog (Taira et al., 1993).
LIM homeobox genes are differentially expressed in primary motoneurons
We examined the embryonic expression of six LIM homeobox genes by in situ RNA hybridization. We did not detect RNA of lim1 (Toyama et al., 1995a), lim5 (Toyama et al., 1995b) or lim6 (R. Toyama and I. B. D., unpublished results) in the ventral spinal cord of 18 hour embryos nor did double labeling experiments of the type described below identify lim6 expression in primary motoneurons (B. A. and J. S. E., unpublished results); therefore, these genes do not appear to be expressed in primary motoneurons prior to axogenesis. Probes for islet1, islet2 and lim3, however, detected expression patterns consistent with the positions of primary motoneurons. We identified the primary motoneurons that express islet1, islet2 and lim3 by injecting fluorescent vital dye into individual cells and, subsequently, probing for gene expression by in situ RNA hybridization. Dye was injected into cells of live 18-22 hour embryos, which were viewed periodically with epifluorescence. Embryos were fixed and hybridized with digoxygenin-labeled RNA probes once motoneuron axons developed their characteristic projections. At the time primary motoneurons develop their axonal projections, additional cells within the spinal cord express islet1, islet2 and lim3 (see below); however, the double-labeling technique allows us to distinguish clearly primary motoneurons from other cell types. The midsegmentally located CaP motoneuron expresses islet2 (Fig. 2A,B) and lim3 RNAs (Fig. 2C,D). VaP, which is closely apposed to CaP and has equivalent developmental potential (Eisen, 1992), also expresses islet2 and lim3 (data not shown). MiP, located near the overlying somite furrow, expresses islet1 (Fig. 2E,F) as well as lim3 (Fig. 2G,H). RoP, which lies just rostral to MiP near the somite border, also expresses islet1 (Fig. 2I,J) and lim3 (Fig. 2K,L). Thus, RoP and MiP can be differentiated from CaP and VaP on the basis of LIM homeobox gene expression. All four cells express lim3 RNA. RoP and MiP express islet1 RNA, but not islet2, while CaP and VaP express islet2 RNA, but not islet1.
Expression of LIM homeobox genes precedes primary motoneuron axogenesis
Expression of LIM homeobox genes is initiated at form a neural keel, which continues to thicken with further condensation of the neural plate where the ventral spinal cord will arise (Papan and Campos-Ortega, 1994). We first detect islet1 RNA near the midline in the early stages of keel formation, when one or two somites are evident (10.3-10.7 hours). islet1 RNA is initially confined to several irregularly spaced cells on both sides of the differentiating floorplate at the level of the first somites to form (Fig. 3A). Expression rapidly expands posteriorly into the neural plate adjacent to unsegmented mesoderm, forming two longitudinal columns of cells (data not shown). islet1 expression proceeds posteriorly at approximately the same rate as somite furrow formation, but maintains a posterior limit of expression about four somite equivalent lengths caudal to the most recently formed somite furrow (data not shown). lim3 expression is first evident at the 3-somite stage (11 hours) in bilateral, discontinuous, longitudinal columns bordering the floorplate. Like islet1, lim3 expression shows no clear segmental pattern at early stages (Fig. 3B). lim3 expression also extends into the unsegmented region of the embryo, but about one half as far as islet1. islet2 expression is first detected at the 7-somite stage (12.5 hours) in bilateral columns bordering the floorplate (data not shown). The posterior extent of islet2 expression roughly corresponds with the most posterior somite.
Patterns of islet1, islet2 and lim3 expression in the spinal cord are very dynamic through early neurogenesis. Therefore, we focus on the patterns at a single axial level, somites 6 and 7, at different developmental stages. The patterns that we describe, although typical, are not invariant as minor differences in the timing of expression are evident in individual embryos. islet2 exhibits the simplest pattern. In 8-somite-stage (13 hour) embryos, islet2 RNA is found in one or two cells per hemisegment, but without a clear spatial arrangement (Fig. 4A). By the 10-somite stage (14 hours), islet2 is still expressed in one or two cells per hemisegment, but they are located in a regular position in the ventral spinal cord, midway between somite borders (hereafter referred to as midsegment cells) (Fig. 4B). This pattern is maintained through the 12-somite (15 hour), 16-somite (17 hour) and 19-somite (18.5 hour) stages early stages of spinal cord development. The zebrafish spinal cord arises from convergent movements of cells within the neuroepithelium toward the dorsal midline of the embryo (Schmitz et al., 1993; Kimmel et al., 1994; Papan and Campos-Ortega, 1994). These movements initially (Fig. 4C,D,E). From our cell labeling analysis described above, we conclude that these cells are CaP and VaP. Additionally, islet2 expression appears in dorsally located cells at the 16-somite and the 19-somite stages (Fig. 4D,E). The positions of these cells are consistent with the distribution of Rohon-Beard (RB) sensory neurons (Bernhardt et al., 1990).
lim3 is expressed in a continuously expanding number of cells. At the 8-somite stage, lim3 RNA is detected in one to three cells near somite borders (border cells) (Fig. 4F). From the 10-somite through 12-somite stages, lim3 is typically expressed in three border cells (Fig. 4G,H). In 16-somite-stage embryos, lim3 expression is detected in several border cells as well as midsegment cells (Fig. 4I). By the 19-somite stage, lim3-expressing cells are broadly distributed in the ventral spinal cord (Fig. 4J).
We have shown that lim3 is expressed in all four primary motoneurons. Thus, two of the three lim3-expressing border cells are likely MiP and RoP. The third cell is probably an interneuron known as VeLD (Bernhardt et al., 1990) (B. A. and J. S. E., unpublished results). The lim3-expressing midsegment cells likely include CaP and VaP. The additional lim3-expressing cells are probably secondary motoneurons, which develop after primary motoneurons (Myers, 1985; Myers et al., 1986).
The pattern of islet1 RNA in the ventral spinal cord undergoes dramatic changes. In 8-somite-stage embryos, islet1 is expressed in two or three unevenly spaced cells per hemisegment (Fig. 4K). By the 10-somite stage, the distribution of islet1-expressing cells is more regular. islet1 RNA is usually found in one border cell and one or two midsegment cells (Fig. 4L). A transition of the islet1 pattern can be seen in 12-somite-stage embryos (Fig. 4M). First, expression in midsegment cells is low or undetectable. Second, expression in border cells at somite levels 7 or 8 is also low or undetectable. In contrast, border cells at both anterior and posterior axial levels exhibit high levels of islet1. Because the islet1 pattern develops in an anterior-to-posterior sequence (see below), this means high levels of islet1 expression are lost and then regained in single border cells within approximately one hour. We are, however, unable to determine if reinitiation of islet1 expression occurs in the same border cell as the original expression, or in a nearby cell. At the 16-somite stage, single border cells located directly below somite furrows express islet1 (Fig. 4N) and in 19-somite-stage embryos, islet1 is expressed in a second border cell adjacent to the first (Fig. 4O). The two border cells that stably express islet1 are MiP and RoP. The position of the first of these two cells to express islet1 RNA relative to the somite furrow indicates that it is MiP. Double islet1 and islet2 RNA labeling experiments show that the transient expression of islet1 in midsegment cells occurs in CaP and VaP (see below). Thus, islet1 RNA is expressed in each of the primary motoneurons during some part of the cell’s history. Like islet2, islet1 RNA is also detected in dorsally located cells that are likely RB neurons.
The islet1 expression pattern reveals an anteroposterior sequence of spinal cord development
In vertebrates, morphogenesis of the trunk occurs in an anterior-to-posterior sequence, as can be seen most easily in the formation of midbody somites. The dynamic nature of islet1 expression allowed us to determine that the development of molecular pattern in the spinal cord occurs in a similar fashion. Fig. 5 shows islet1 expression at different anteroposterior positions in a 19-somite-stage embryo; the anteroposterior pattern reflects the patterns observed at a single axial level during development (compare to Fig. 4K-O). islet1-expressing cells are closely spaced in the spinal cord coincident with unsegmented mesoderm and newly formed somites (Fig. 5A). At the 15-to 16-somite level, islet1 RNA can be seen in one border cell and at low levels in one or two midsegment cells (Fig. 5B). At somite levels 13-14, expression at the somite 14 border cell declines, presumably transiently, and midsegment cell expression remains at low levels (Fig. 5C). At somite levels 10-11, single cells at the somite boundaries express islet1 RNA and midsegment cell expression is absent (Fig. 5D). Two border cells express islet1 at somite level 8-9 (Fig. 5E) and, at somite levels 5-6, a cluster of about four cells express islet1 near the somite boundaries (Fig. 5F). As development proceeds, additional cells in the ventral spinal cord express islet1 (data not shown). In addition to MiP and RoP, this growing cluster of islet1-expressing cells likely includes a subset of secondary motoneurons.
Although the progression of the islet1 pattern is played out along the anteroposterior axis of the trunk, the initial pattern in the posterior trunk and tail differs from the initial islet1 pattern in the most anterior spinal segments. From the onset of expression in the anterior regions of the spinal cord, islet1 RNA appears in cells separated by cells in which islet1 is not detected. In the posterior trunk, islet1-positive cells are more closely spaced, often being contiguous (compare Fig. 3A and Fig. 5A).
islet1 RNA is transiently expressed in CaP and VaP motoneurons
The position of the islet1-positive cells midway between somite borders in young segments is indicative of CaP and VaP motoneurons. As islet2 expression is a stable marker of CaP and VaP fates, we asked whether CaP and VaP transiently express islet1 RNA by simultaneously probing 21-somite-stage embryos for both gene products. At the level of somite 14, islet1 and islet2 probes label different cells (Fig. 6A,B). In contrast, in more posterior regions, somite 17 for example, islet1 and islet2 RNAs are coexpressed in midsegment cells (Fig. 6C,D). Thus, the initial expression of islet1 includes CaP and VaP. Downregulation of islet1 RNA in CaP and VaP, as indicated by decreased staining intensity, occurs within 1-2 hours after islet2 is first detected in cells at a comparable anteroposterior level.
Positional determination of gene expression and primary motoneuronal fate
The intricate patterns of gene expression that we have described indicate that transcription of islet1, islet2 and lim3 is highly regulated. We predict that these patterns arise as a result of positional information that determines in which cell each gene is transcribed, but it is unclear how such information might be distributed within the spinal cord and when it is utilized by a cell to determine a program of LIM homeobox gene expression. To explore these issues, we transplanted individual primary motoneurons to new positions at different developmental stages and asked whether position influenced LIM homeobox gene expression. We chose to transplant cells from the MiP position of dye-labeled donor embryos to various positions of unlabeled hosts and probe for islet2 since it is expressed in CaP and VaP, and not in MiP, through at least 30 hours (B. A. and J. S. E., unpublished results). Thus, islet2 expression in the transplanted cell would indicate that its fate had been altered. When cells were transplanted from the MiP position of a donor to the MiP position of a host, they developed a MiP morphology and did not express islet2 (Table 1). When cells were transplanted from the MiP position to the CaP position within approximately 1 hour before the initiation of axogenesis, they maintained a MiP fate and, also, did not express islet2 (Fig. 7A; Table 1). In contrast, cells transplanted from the MiP position to the CaP position 2-3.5 hours before MiP axogenesis adopted the CaP fate. All of these cells expressed islet2 (Fig. 7B; Table 1). These experiments demonstrate that the transcriptional state of islet2 is highly sensitive to cell position within the spinal cord and does not become fixed until shortly before axogenesis. Furthermore, the exact correlation of islet2 expression with CaP, but not MiP identity, shows that islet2 expression is tightly coupled to cell fate.
DISCUSSION
LIM homeobox genes and motoneuronal identity
Specification of motoneuronal identity should be reflected in patterns of gene expression that uniquely demarcate motoneurons prior to their overt differentiation. We have shown that this is the case for zebrafish primary motoneurons, each of which express a particular temporal sequence of LIM homeobox genes. Initiation of LIM homeobox gene expression precedes overt differentiation of primary motoneurons, marked by axogenesis, by several hours. The patterns of gene expression, although initially dynamic, are resolved before axogenesis so that distinct primary motoneurons express specific combinations of LIM homeobox genes. Thus, the expression patterns of LIM homeobox genes show that primary motoneurons are specified before the initiation of axonal outgrowth.
LIM homeobox genes may have a role in motoneuronal fate specification that is widely conserved among vertebrates. Tsuchida et al. (1994) isolated a number of chick LIM homeobox genes and found correlations between different combinations of gene products and the columnar organization of motoneurons. Some, but not all, features of LIM homeobox gene expression are shared between chick motoneurons and zebrafish primary motoneurons. First, zebrafish primary motoneurons express three of four LIM homeobox gene homologs expressed in chick motoneurons; only Lim-1 is expressed in chick motoneurons and not zebrafish primary motoneurons. In chick, Lim-1 expression is restricted to motoneurons of the lateral subdivision of the lateral motor column (LMCl), which innervate limb muscle. Zebrafish primary motoneurons do not innervate limb muscle, suggesting that Lim-1 may have a distinct role in specifying limb-specific motoneurons. Second, all zebrafish primary motoneurons express lim3. In chick, Lim-3 expression is limited to motoneurons of the medial subdivision of the median motor column (MMCm). MMCm motoneurons and zebrafish primary motoneurons both innervate axial muscle, suggesting that lim3 may be particularly involved in specifying motoneurons that innervate this muscle type. Third, all chick motoneurons express Islet-1 and Islet-2 at some stage of their differentiation. This is not the case for all zebrafish primary motoneurons as islet2 is never expressed in MiP and RoP. Finally, the order in which these genes are expressed differs somewhat between zebrafish primary motoneurons and chick motoneurons. In all chick motoneurons, Islet-1 RNA is detected before RNA of other LIM homeobox genes. The same pattern is observed in zebrafish CaP, VaP and MiP. However, our results raise the possibility that lim3 expression precedes that of islet1 in RoP. Thus, although primary motoneurons appear to employ the same set of LIM homeobox genes as chick motoneurons, there are significant differences in combinatorial expression and timing, perhaps identifying fundamental differences between specification of zebrafish primary motoneurons and chick motoneurons. It will be necessary to characterize LIM homeobox gene expression in secondary motoneurons of zebrafish, which innervate axial and pectoral fin muscle, as well as in motoneurons of other vertebrate species to identify essential features of a putative combinatorial code of LIM homeodomain proteins involved in motoneuronal fate specification
The set of transcription factors encoded by the islet1, islet2 and lim3 genes cannot account for the full range of motoneuronal identities evident within chick spinal cord nor among primary motoneurons in zebrafish. For example, distinct motor pools that lie at characteristic positions within the chick MMCm and innervate either dorsal or ventral axial muscles (Gutman et al., 1993) are not identified by differential expression of LIM homeobox genes (Tsuchida et al., 1994). Similarly, zebrafish RoP cannot be differentiated from MiP, nor CaP from VaP on the basis of LIM homeobox gene expression at the time of axogenesis. What, then, might account for individual fate differences? Ablation and transplantation experiments showed that CaP and VaP have equivalent developmental potential, which is maintained nearly until the time of VaP death (Eisen, 1992). Differentiation of CaP and VaP fates, therefore, may not result from the patterned, differential expression of transcription factors controlling sets of downstream genes. Instead, we suspect that CaP and VaP fate differentiation results from the initiation of a cell death pathway resulting from an interaction between CaP and VaP. RoP maintains its identity after MiP ablation (Pike and Eisen, 1990), as does MiP after RoP ablation (J. S. E., unpublished results) showing, in contrast to CaP and VaP, that MiP and RoP do not constitute an equivalence pair. MiP and RoP could be differentiated by the timing of islet1 expression relative to morphological differentiation. Approximately 3 hours separates the time at which we can first identify islet1 expression in a single border cell and expression in two cells in that region, therefore, it is possible that early islet1 expression determines one cell fate, and late expression, another. Alternatively, an unidentified, differentially expressed gene or set of genes, acting in combination with lim3 and islet1, may specify the individual MiP and RoP fates. Such a function may be provided by additional genes of the LIM homeobox class or by transcription factors of another class. The differential expression of a LIM domain binding partner that affects the activity of a LIM homeodomain protein could, also, determine a distinction between MiP and RoP. For example, DNA-binding specificity is enhanced by a cooperative interaction of the LIM homeodomain protein MEC-3 and the POU homeodomain protein UNC-86 as a heterodimer (Xue et al., 1993). LIM domains act as negative regulatory elements of Xlim-1 function in experiments assaying induction of neural and muscle tissue in X. laevis suggesting a means for the modification of LIM homeodomain protein function (Taira et al., 1994).
Patterned LIM homeobox gene expression and commitment of cell fate
Single cell transplantation experiments have shown that zebrafish primary motoneurons become committed to particular fates after their terminal division, but before morphological differentiation (Eisen, 1991). How does commitment correlate with expression of LIM homeobox genes? Expression of each of the LIM homeobox genes described here is initiated in CaP and VaP several hours before commitment. The event that correlates best with the time of commitment to the CaP fate is the loss of islet1 RNA. In trunk segments, CaP becomes committed to its fate by 16 hours (Eisen, 1991). Detectable amounts of islet1 RNA are nearly absent in CaP and VaP of trunk segments by 15 hours. Thus, islet1 expression could prevent a cell from adopting the CaP fate. MiP becomes committed by about 17-18 hours, which follows the onset of islet1 and lim3 expression in trunk segments. The expression of particular combinations of islet1, islet2 and lim3 in primary motoneurons before commitment and before axonal development suggests these genes may contribute to primary motoneuron specification.
By transplanting primary motoneurons and probing for gene expression, we have demonstrated a tight correlation between commitment to a particular motoneuronal fate and LIM homeobox gene expression. Cells fated to become MiP, when transplanted to the CaP position at least 2 hours prior to axogenesis, developed as CaP or VaP and initiated islet2 expression. When the same type of transplant was done within 1 hour of axogenesis, the transplanted cell maintained a MiP fate and did not express islet2. These results support a possible role for islet2 in the specification of CaP and VaP fates.
Perhaps most importantly, the transplantation experiments provide insight to the patterning mechanisms that must exist within the spinal cord. Previous studies have shown that notochord and floorplate can induce the expression of motoneuronal markers in spinal cord (Yamada et al., 1991, 1993). In zebrafish, the onset of islet1 expression occurs in cells close to the differentiating notochord and floorplate. Thus, the time and place of islet1 expression in zebrafish is consistent with induction by a signal originating in the notochord and/or floorplate. How an axial inductive signal is converted to the patterned distribution of different cell types is considerably less clear. We have shown that transplantation of a cell to a position only a few cell diameters away can change both gene expression and cell fate. Thus, a finely grained system of positional information, to which transcriptional control of LIM homeobox genes is responsive, must exist within the spinal cord. It will be necessary to determine how patterned gene expression is established and maintained to understand how diverse cell types are generated in the vertebrate central nervous system.
ACKNOWLEDGEMENTS
We thank Christine Beattie, Bruce Bowerman, Robert Kelsh, Charles Kimmel, John Postlethwait and David Raible for comments on earlier versions of the manuscript. We are indebted to Barbara Jones and Martin Petkovich for the genomic library, Bob Riggleman, Kathryn Helde and David Grunwald for the cDNA library, Inna Korzh-Sleptsova, Helena Alstermark and Maria Lind for technical help, and Masanori Taira and Reiko Toyama for advice. Supported by: NFR and MFR (Sweden) (T. E.); NS23915, NS01476 (J. S. E.); F32HD07658 (B. A.).